(PDF) Thinking, fast and slow by daniel kahneman - PDFSLIDE.US (2024)

(PDF) Thinking, fast and slow by daniel kahneman - PDFSLIDE.US (1)

(PDF) Thinking, fast and slow by daniel kahneman - PDFSLIDE.US (2)

(PDF) Thinking, fast and slow by daniel kahneman - PDFSLIDE.US (3)

In memory of Amos Tversky

(PDF) Thinking, fast and slow by daniel kahneman - PDFSLIDE.US (4)

Contents

Introduction

Part I. Two Systems

1. The Characters of the Story

2. Attention and Effort

3. The Lazy Controller

4. The Associative Machine

5. Cognitive Ease

6. Norms, Surprises, and Causes

7. A Machine for Jumping to Conclusions

8. How Judgments Happen

9. Answering an Easier Question

Part II. Heuristics and Biases

10. The Law of Small Numbers

<5>11. Anchors

12. The Science of Availability

13. Availability, Emotion, and Risk

14. Tom W’s Specialty

(PDF) Thinking, fast and slow by daniel kahneman - PDFSLIDE.US (5)

15. Linda: Less is More

16. Causes Trump Statistics

17. Regression to the Mean

18. Taming Intuitive Predictions

Part III. Overconfidence

19. The Illusion of Understanding

20. The Illusion of Validity

21. Intuitions Vs. Formulas

22. Expert Intuition: When Can We Trust It?

23. The Outside View

24. The Engine of Capitalism

Part IV. Choices25. Bernoulli’s Errors

26. Prospect Theory

27. The Endowment Effect

28. Bad Events

29. The Fourfold Pattern

30. Rare Events

31. Risk Policies

(PDF) Thinking, fast and slow by daniel kahneman - PDFSLIDE.US (6)

32. Keeping Score

33. Reversals

34. Frames and Reality

Part V. Two Selves

35. Two Selves

36. Life as a Story

37. Experienced Well-Being

38. Thinking About Life

Conclusions

Appendix A: Judgment UnderUncertainty

Appendix B: Choices, Values, and Frames

Acknowledgments

Notes

Index

(PDF) Thinking, fast and slow by daniel kahneman - PDFSLIDE.US (7)

Introduction

Every author, I suppose, has in mind a setting in which readers of his or herwork could benefit from having read it. Mine is the proverbial officewatercooler, where opinions are shared and gossip is exchanged. I hopeto enrich the vocabulary that people use when they talk about thejudgments and choices of others, the company’s new policies, or acolleague’s investment decisions. Why be concerned with gossip?Because it is much easier, as well as far more enjoyable, to identify andlabel the mistakes of others than to recognize our own. Questioning whatwe believe and want is difficult at the best of times, and especially difficultwhen we most need to do it, but we can benefit from the informed opinionsof others. Many of us spontaneously anticipate how friends and colleagueswill evaluate our choices; the quality and content of these anticipatedjudgments therefore matters. The expectation of intelligent gossip is apowerful motive for serious self-criticism, more powerful than New Yearresolutions to improve one’s decision making at work and at home.

To be a good diagnostician, a physician needs to acquire a large set oflabels for diseases, each of which binds an idea of the illness and itssymptoms, possible antecedents and causes, possible developments andconsequences, and possible interventions to cure or mitigate the illness.Learning medicine consists in part of learning the language of medicine. Adeeper understanding of judgments and choices also requires a richervocabulary than is available in everyday language. The hope for informedgossip is that there are distinctive patterns in the errors people make.Systematic errors are known as biases, and they recur predictably inparticular circumstances. When the handsome and confident speakerbounds onto the stage, for example, you can anticipate that the audiencewill judge his comments more favorably than he deserves. The availabilityof a diagnostic label for this bias—the halo effect—makes it easier toanticipate, recognize, and understand.

When you are asked what you are thinking about, you can normallyanswer. You believe you know what goes on in your mind, which oftenconsists of one conscious thought leading in an orderly way to another. Butthat is not the only way the mind works, nor indeed is that the typical way.Most impressions and thoughts arise in your conscious experience withoutyour knowing how they got there. You cannot tracryd>e how you came tothe belief that there is a lamp on the desk in front of you, or how youdetected a hint of irritation in your spouse’s voice on the telephone, or how

(PDF) Thinking, fast and slow by daniel kahneman - PDFSLIDE.US (8)

you managed to avoid a threat on the road before you became consciouslyaware of it. The mental work that produces impressions, intuitions, andmany decisions goes on in silence in our mind.

Much of the discussion in this book is about biases of intuition. However,the focus on error does not denigrate human intelligence, any more thanthe attention to diseases in medical texts denies good health. Most of usare healthy most of the time, and most of our judgments and actions areappropriate most of the time. As we navigate our lives, we normally allowourselves to be guided by impressions and feelings, and the confidencewe have in our intuitive beliefs and preferences is usually justified. But notalways. We are often confident even when we are wrong, and an objectiveobserver is more likely to detect our errors than we are.

So this is my aim for watercooler conversations: improve the ability toidentify and understand errors of judgment and choice, in others andeventually in ourselves, by providing a richer and more precise language todiscuss them. In at least some cases, an accurate diagnosis may suggestan intervention to limit the damage that bad judgments and choices oftencause.

Origins

This book presents my current understanding of judgment and decisionmaking, which has been shaped by psychological discoveries of recentdecades. However, I trace the central ideas to the lucky day in 1969 when Iasked a colleague to speak as a guest to a seminar I was teaching in theDepartment of Psychology at the Hebrew University of Jerusalem. AmosTversky was considered a rising star in the field of decision research—indeed, in anything he did—so I knew we would have an interesting time.Many people who knew Amos thought he was the most intelligent personthey had ever met. He was brilliant, voluble, and charismatic. He was alsoblessed with a perfect memory for jokes and an exceptional ability to usethem to make a point. There was never a dull moment when Amos wasaround. He was then thirty-two; I was thirty-five.

Amos told the class about an ongoing program of research at theUniversity of Michigan that sought to answer this question: Are peoplegood intuitive statisticians? We already knew that people are goodintuitive grammarians: at age four a child effortlessly conforms to the rulesof grammar as she speaks, although she has no idea that such rules exist.Do people have a similar intuitive feel for the basic principles of statistics?Amos reported that the answer was a qualified yes. We had a lively debatein the seminar and ultimately concluded that a qualified no was a better

(PDF) Thinking, fast and slow by daniel kahneman - PDFSLIDE.US (9)

answer.Amos and I enjoyed the exchange and concluded that intuitive statistics

was an interesting topic and that it would be fun to explore it together. ThatFriday we met for lunch at Café Rimon, the favorite hangout of bohemiansand professors in Jerusalem, and planned a study of the statisticalintuitions of sophisticated researchers. We had concluded in the seminarthat our own intuitions were deficient. In spite of years of teaching andusing statistics, we had not developed an intuitive sense of the reliability ofstatistical results observed in small samples. Our subjective judgmentswere biased: we were far too willing to believe research findings based oninadequate evidence and prone to collect too few observations in our ownresearch. The goal of our study was to examine whether other researcherssuffered from the same affliction.

We prepared a survey that included realistic scenarios of statisticalissues that arise in research. Amos collected the responses of a group ofexpert participants in a meeting of the Society of MathematicalPsychology, including the authors of two statistical textbooks. As expected,we found that our expert colleagues, like us, greatly exaggerated thelikelihood that the original result of an experiment would be successfullyreplicated even with a small sample. They also gave very poor advice to afictitious graduate student about the number of observations she neededto collect. Even statisticians were not good intuitive statisticians.

While writing the article that reported these findings, Amos and Idiscovered that we enjoyed working together. Amos was always veryfunny, and in his presence I became funny as well, so we spent hours ofsolid work in continuous amusement. The pleasure we found in workingtogether made us exceptionally patient; it is much easier to strive forperfection when you are never bored. Perhaps most important, wechecked our critical weapons at the door. Both Amos and I were criticaland argumentative, he even more than I, but during the years of ourcollaboration neither of us ever rejected out of hand anything the othersaid. Indeed, one of the great joys I found in the collaboration was thatAmos frequently saw the point of my vague ideas much more clearly than Idid. Amos was the more logical thinker, with an orientation to theory andan unfailing sense of direction. I was more intuitive and rooted in thepsychology of perception, from which we borrowed many ideas. We weresufficiently similar to understand each other easily, and sufficiently differentto surprise each other. We developed a routine in which we spent much ofour working days together, often on long walks. For the next fourteen yearsour collaboration was the focus of our lives, and the work we did togetherduring those years was the best either of us ever did.

We quickly adopted a practice that we maintained for many years. Our

(PDF) Thinking, fast and slow by daniel kahneman - PDFSLIDE.US (10)

research was a conversation, in which we invented questions and jointlyexamined our intuitive answers. Each question was a small experiment,and we carried out many experiments in a single day. We were notseriously looking for the correct answer to the statistical questions weposed. Our aim was to identify and analyze the intuitive answer, the firstone that came to mind, the one we were tempted to make even when weknew it to be wrong. We believed—correctly, as it happened—that anyintuition that the two of us shared would be shared by many other peopleas well, and that it would be easy to demonstrate its effects on judgments.

We once discovered with great delight that we had identical silly ideasabout the future professions of several toddlers we both knew. We couldidentify the argumentative three-year-old lawyer, the nerdy professor, theempathetic and mildly intrusive psychotherapist. Of course thesepredictions were absurd, but we still found them appealing. It was alsoclear that our intuitions were governed by the resemblance of each child tothe cultural stereotype of a profession. The amusing exercise helped usdevelop a theory that was emerging in our minds at the time, about the roleof resemblance in predictions. We went on to test and elaborate thattheory in dozens of experiments, as in the following example.

As you consider the next question, please assume that Steve wasselected at random from a representative sample:

An individual has been described by a neighbor as follows:“Steve is very shy and withdrawn, invariably helpful but with littleinterest in people or in the world of reality. A meek and tidy soul,he has a need for order and structurut and stre, and a passion fordetail.” Is Steve more likely to be a librarian or a farmer?

The resemblance of Steve’s personality to that of a stereotypical librarianstrikes everyone immediately, but equally relevant statisticalconsiderations are almost always ignored. Did it occur to you that thereare more than 20 male farmers for each male librarian in the UnitedStates? Because there are so many more farmers, it is almost certain thatmore “meek and tidy” souls will be found on tractors than at libraryinformation desks. However, we found that participants in our experimentsignored the relevant statistical facts and relied exclusively on resemblance.We proposed that they used resemblance as a simplifying heuristic(roughly, a rule of thumb) to make a difficult judgment. The reliance on theheuristic caused predictable biases (systematic errors) in theirpredictions.

On another occasion, Amos and I wondered about the rate of divorceamong professors in our university. We noticed that the question triggered

(PDF) Thinking, fast and slow by daniel kahneman - PDFSLIDE.US (11)

a search of memory for divorced professors we knew or knew about, andthat we judged the size of categories by the ease with which instancescame to mind. We called this reliance on the ease of memory search theavailability heuristic. In one of our studies, we asked participants to answera simple question about words in a typical English text:

Consider the letter K.Is K more likely to appear as the first letter in a word OR as thethird letter?

As any Scrabble player knows, it is much easier to come up with wordsthat begin with a particular letter than to find words that have the sameletter in the third position. This is true for every letter of the alphabet. Wetherefore expected respondents to exaggerate the frequency of lettersappearing in the first position—even those letters (such as K, L, N, R, V)which in fact occur more frequently in the third position. Here again, thereliance on a heuristic produces a predictable bias in judgments. Forexample, I recently came to doubt my long-held impression that adultery ismore common among politicians than among physicians or lawyers. I hadeven come up with explanations for that “fact,” including the aphrodisiaceffect of power and the temptations of life away from home. I eventuallyrealized that the transgressions of politicians are much more likely to bereported than the transgressions of lawyers and doctors. My intuitiveimpression could be due entirely to journalists’ choices of topics and to myreliance on the availability heuristic.

Amos and I spent several years studying and documenting biases ofintuitive thinking in various tasks—assigning probabilities to events,forecasting the future, assessing hypotheses, and estimating frequencies.In the fifth year of our collaboration, we presented our main findings inScience magazine, a publication read by scholars in many disciplines. Thearticle (which is reproduced in full at the end of this book) was titled“Judgment Under Uncertainty: Heuristics and Biases.” It described thesimplifying shortcuts of intuitive thinking and explained some 20 biases asmanifestations of these heuristics—and also as demonstrations of the roleof heuristics in judgment.

Historians of science have often noted that at any given time scholars ina particular field tend to share basic re share assumptions about theirsubject. Social scientists are no exception; they rely on a view of humannature that provides the background of most discussions of specificbehaviors but is rarely questioned. Social scientists in the 1970s broadlyaccepted two ideas about human nature. First, people are generally

(PDF) Thinking, fast and slow by daniel kahneman - PDFSLIDE.US (12)

rational, and their thinking is normally sound. Second, emotions such asfear, affection, and hatred explain most of the occasions on which peopledepart from rationality. Our article challenged both assumptions withoutdiscussing them directly. We documented systematic errors in the thinkingof normal people, and we traced these errors to the design of themachinery of cognition rather than to the corruption of thought by emotion.

Our article attracted much more attention than we had expected, and itremains one of the most highly cited works in social science (more thanthree hundred scholarly articles referred to it in 2010). Scholars in otherdisciplines found it useful, and the ideas of heuristics and biases havebeen used productively in many fields, including medical diagnosis, legaljudgment, intelligence analysis, philosophy, finance, statistics, and militarystrategy.

For example, students of policy have noted that the availability heuristichelps explain why some issues are highly salient in the public’s mind whileothers are neglected. People tend to assess the relative importance ofissues by the ease with which they are retrieved from memory—and this islargely determined by the extent of coverage in the media. Frequentlymentioned topics populate the mind even as others slip away fromawareness. In turn, what the media choose to report corresponds to theirview of what is currently on the public’s mind. It is no accident thatauthoritarian regimes exert substantial pressure on independent media.Because public interest is most easily aroused by dramatic events and bycelebrities, media feeding frenzies are common. For several weeks afterMichael Jackson’s death, for example, it was virtually impossible to find atelevision channel reporting on another topic. In contrast, there is littlecoverage of critical but unexciting issues that provide less drama, such asdeclining educational standards or overinvestment of medical resources inthe last year of life. (As I write this, I notice that my choice of “little-covered”examples was guided by availability. The topics I chose as examples arementioned often; equally important issues that are less available did notcome to my mind.)

We did not fully realize it at the time, but a key reason for the broadappeal of “heuristics and biases” outside psychology was an incidentalfeature of our work: we almost always included in our articles the full text ofthe questions we had asked ourselves and our respondents. Thesequestions served as demonstrations for the reader, allowing him torecognize how his own thinking was tripped up by cognitive biases. I hopeyou had such an experience as you read the question about Steve thelibrarian, which was intended to help you appreciate the power ofresemblance as a cue to probability and to see how easy it is to ignorerelevant statistical facts.

(PDF) Thinking, fast and slow by daniel kahneman - PDFSLIDE.US (13)

The use of demonstrations provided scholars from diverse disciplines—notably philosophers and economists—an unusual opportunity to observepossible flaws in their own thinking. Having seen themselves fail, theybecame more likely to question the dogmatic assumption, prevalent at thetime, that the human mind is rational and logical. The choice of methodwas crucial: if we had reported results of only conventional experiments,the article would have been less noteworthy and less memorable.Furthermore, skeptical readers would have distanced themselves from theresults by attributing judgment errors to the familiar l the famifecklessnessof undergraduates, the typical participants in psychological studies. Ofcourse, we did not choose demonstrations over standard experimentsbecause we wanted to influence philosophers and economists. Wepreferred demonstrations because they were more fun, and we were luckyin our choice of method as well as in many other ways. A recurrent themeof this book is that luck plays a large role in every story of success; it isalmost always easy to identify a small change in the story that would haveturned a remarkable achievement into a mediocre outcome. Our story wasno exception.

The reaction to our work was not uniformly positive. In particular, ourfocus on biases was criticized as suggesting an unfairly negative view ofthe mind. As expected in normal science, some investigators refined ourideas and others offered plausible alternatives. By and large, though, theidea that our minds are susceptible to systematic errors is now generallyaccepted. Our research on judgment had far more effect on social sciencethan we thought possible when we were working on it.

Immediately after completing our review of judgment, we switched ourattention to decision making under uncertainty. Our goal was to develop apsychological theory of how people make decisions about simplegambles. For example: Would you accept a bet on the toss of a coin whereyou win $130 if the coin shows heads and lose $100 if it shows tails?These elementary choices had long been used to examine broadquestions about decision making, such as the relative weight that peopleassign to sure things and to uncertain outcomes. Our method did notchange: we spent many days making up choice problems and examiningwhether our intuitive preferences conformed to the logic of choice. Hereagain, as in judgment, we observed systematic biases in our owndecisions, intuitive preferences that consistently violated the rules ofrational choice. Five years after the Science article, we published“Prospect Theory: An Analysis of Decision Under Risk,” a theory of choicethat is by some counts more influential than our work on judgment, and isone of the foundations of behavioral economics.

(PDF) Thinking, fast and slow by daniel kahneman - PDFSLIDE.US (14)

Until geographical separation made it too difficult to go on, Amos and Ienjoyed the extraordinary good fortune of a shared mind that was superiorto our individual minds and of a relationship that made our work fun as wellas productive. Our collaboration on judgment and decision making was thereason for the Nobel Prize that I received in 2002, which Amos would haveshared had he not died, aged fifty-nine, in 1996.

Where we are now

This book is not intended as an exposition of the early research that Amosand I conducted together, a task that has been ably carried out by manyauthors over the years. My main aim here is to present a view of how themind works that draws on recent developments in cognitive and socialpsychology. One of the more important developments is that we nowunderstand the marvels as well as the flaws of intuitive thought.

Amos and I did not address accurate intuitions beyond the casualstatement that judgment heuristics “are quite useful, but sometimes lead tosevere and systematic errors.” We focused on biases, both because wefound them interesting in their own right and because they providedevidence for the heuristics of judgment. We did not ask ourselves whetherall intuitive judgments under uncertainty are produced by the heuristics westudied; it is now clear that they are not. In particular, the accurate intuitionsof experts are better explained by the effects of prolonged practice than byheuristics. We can now draw a richer andigha riche more balancedpicture, in which skill and heuristics are alternative sources of intuitivejudgments and choices.

The psychologist Gary Klein tells the story of a team of firefighters thatentered a house in which the kitchen was on fire. Soon after they startedhosing down the kitchen, the commander heard himself shout, “Let’s getout of here!” without realizing why. The floor collapsed almost immediatelyafter the firefighters escaped. Only after the fact did the commander realizethat the fire had been unusually quiet and that his ears had been unusuallyhot. Together, these impressions prompted what he called a “sixth senseof danger.” He had no idea what was wrong, but he knew something waswrong. It turned out that the heart of the fire had not been in the kitchen butin the basement beneath where the men had stood.

We have all heard such stories of expert intuition: the chess master whowalks past a street game and announces “White mates in three” withoutstopping, or the physician who makes a complex diagnosis after a singleglance at a patient. Expert intuition strikes us as magical, but it is not.Indeed, each of us performs feats of intuitive expertise many times each

(PDF) Thinking, fast and slow by daniel kahneman - PDFSLIDE.US (15)

day. Most of us are pitch-perfect in detecting anger in the first word of atelephone call, recognize as we enter a room that we were the subject ofthe conversation, and quickly react to subtle signs that the driver of the carin the next lane is dangerous. Our everyday intuitive abilities are no lessmarvelous than the striking insights of an experienced firefighter orphysician—only more common.

The psychology of accurate intuition involves no magic. Perhaps thebest short statement of it is by the great Herbert Simon, who studied chessmasters and showed that after thousands of hours of practice they come tosee the pieces on the board differently from the rest of us. You can feelSimon’s impatience with the mythologizing of expert intuition when hewrites: “The situation has provided a cue; this cue has given the expertaccess to information stored in memory, and the information provides theanswer. Intuition is nothing more and nothing less than recognition.”

We are not surprised when a two-year-old looks at a dog and says“doggie!” because we are used to the miracle of children learning torecognize and name things. Simon’s point is that the miracles of expertintuition have the same character. Valid intuitions develop when expertshave learned to recognize familiar elements in a new situation and to act ina manner that is appropriate to it. Good intuitive judgments come to mindwith the same immediacy as “doggie!”

Unfortunately, professionals’ intuitions do not all arise from trueexpertise. Many years ago I visited the chief investment officer of a largefinancial firm, who told me that he had just invested some tens of millions ofdollars in the stock of Ford Motor Company. When I asked how he hadmade that decision, he replied that he had recently attended an automobileshow and had been impressed. “Boy, do they know how to make a car!”was his explanation. He made it very clear that he trusted his gut feelingand was satisfied with himself and with his decision. I found it remarkablethat he had apparently not considered the one question that an economistwould call relevant: Is Ford stock currently underpriced? Instead, he hadlistened to his intuition; he liked the cars, he liked the company, and heliked the idea of owning its stock. From what we know about the accuracyof stock picking, it is reasonable to believe that he did not know what hewas doing.

The specific heuristics that Amos and I studied proviheitudied de littlehelp in understanding how the executive came to invest in Ford stock, but abroader conception of heuristics now exists, which offers a good account.An important advance is that emotion now looms much larger in ourunderstanding of intuitive judgments and choices than it did in the past.The executive’s decision would today be described as an example of theaffect heuristic, where judgments and decisions are guided directly by

(PDF) Thinking, fast and slow by daniel kahneman - PDFSLIDE.US (16)

feelings of liking and disliking, with little deliberation or reasoning.When confronted with a problem—choosing a chess move or deciding

whether to invest in a stock—the machinery of intuitive thought does thebest it can. If the individual has relevant expertise, she will recognize thesituation, and the intuitive solution that comes to her mind is likely to becorrect. This is what happens when a chess master looks at a complexposition: the few moves that immediately occur to him are all strong. Whenthe question is difficult and a skilled solution is not available, intuition stillhas a shot: an answer may come to mind quickly—but it is not an answerto the original question. The question that the executive faced (should Iinvest in Ford stock?) was difficult, but the answer to an easier and relatedquestion (do I like Ford cars?) came readily to his mind and determinedhis choice. This is the essence of intuitive heuristics: when faced with adifficult question, we often answer an easier one instead, usually withoutnoticing the substitution.

The spontaneous search for an intuitive solution sometimes fails—neither an expert solution nor a heuristic answer comes to mind. In suchcases we often find ourselves switching to a slower, more deliberate andeffortful form of thinking. This is the slow thinking of the title. Fast thinkingincludes both variants of intuitive thought—the expert and the heuristic—aswell as the entirely automatic mental activities of perception and memory,the operations that enable you to know there is a lamp on your desk orretrieve the name of the capital of Russia.

The distinction between fast and slow thinking has been explored bymany psychologists over the last twenty-five years. For reasons that Iexplain more fully in the next chapter, I describe mental life by the metaphorof two agents, called System 1 and System 2, which respectively producefast and slow thinking. I speak of the features of intuitive and deliberatethought as if they were traits and dispositions of two characters in yourmind. In the picture that emerges from recent research, the intuitive System1 is more influential than your experience tells you, and it is the secretauthor of many of the choices and judgments you make. Most of this bookis about the workings of System 1 and the mutual influences between itand System 2.

What Comes Next

The book is divided into five parts. Part 1 presents the basic elements of atwo-systems approach to judgment and choice. It elaborates the distinctionbetween the automatic operations of System 1 and the controlledoperations of System 2, and shows how associative memory, the core of

(PDF) Thinking, fast and slow by daniel kahneman - PDFSLIDE.US (17)

System 1, continually constructs a coherent interpretation of what is goingon in our world at any instant. I attempt to give a sense of the complexityand richness of the automatic and often unconscious processes thatunderlie intuitive thinking, and of how these automatic processes explainthe heuristics of judgment. A goal is to introduce a language for thinkingand talking about the mind.

Part 2 updates the study of judgment heuristics and explores a majorpuzzle: Why is it so difficult for us to think statistically? We easily thinkassociativelm 1associay, we think metaphorically, we think causally, butstatistics requires thinking about many things at once, which is somethingthat System 1 is not designed to do.

The difficulties of statistical thinking contribute to the main theme of Part3, which describes a puzzling limitation of our mind: our excessiveconfidence in what we believe we know, and our apparent inability toacknowledge the full extent of our ignorance and the uncertainty of theworld we live in. We are prone to overestimate how much we understandabout the world and to underestimate the role of chance in events.Overconfidence is fed by the illusory certainty of hindsight. My views on thistopic have been influenced by Nassim Taleb, the author of The BlackSwan. I hope for watercooler conversations that intelligently explore thelessons that can be learned from the past while resisting the lure ofhindsight and the illusion of certainty.

The focus of part 4 is a conversation with the discipline of economics onthe nature of decision making and on the assumption that economicagents are rational. This section of the book provides a current view,informed by the two-system model, of the key concepts of prospect theory,the model of choice that Amos and I published in 1979. Subsequentchapters address several ways human choices deviate from the rules ofrationality. I deal with the unfortunate tendency to treat problems inisolation, and with framing effects, where decisions are shaped byinconsequential features of choice problems. These observations, whichare readily explained by the features of System 1, present a deepchallenge to the rationality assumption favored in standard economics.

Part 5 describes recent research that has introduced a distinctionbetween two selves, the experiencing self and the remembering self, whichdo not have the same interests. For example, we can expose people totwo painful experiences. One of these experiences is strictly worse thanthe other, because it is longer. But the automatic formation of memories—a feature of System 1—has its rules, which we can exploit so that theworse episode leaves a better memory. When people later choose whichepisode to repeat, they are, naturally, guided by their remembering self

(PDF) Thinking, fast and slow by daniel kahneman - PDFSLIDE.US (18)

and expose themselves (their experiencing self) to unnecessary pain. Thedistinction between two selves is applied to the measurement of well-being, where we find again that what makes the experiencing self happy isnot quite the same as what satisfies the remembering self. How two selveswithin a single body can pursue happiness raises some difficult questions,both for individuals and for societies that view the well-being of thepopulation as a policy objective.

A concluding chapter explores, in reverse order, the implications of threedistinctions drawn in the book: between the experiencing and theremembering selves, between the conception of agents in classicaleconomics and in behavioral economics (which borrows from psychology),and between the automatic System 1 and the effortful System 2. I return tothe virtues of educating gossip and to what organizations might do toimprove the quality of judgments and decisions that are made on theirbehalf.

Two articles I wrote with Amos are reproduced as appendixes to thebook. The first is the review of judgment under uncertainty that I describedearlier. The second, published in 1984, summarizes prospect theory aswell as our studies of framing effects. The articles present the contributionsthat were cited by the Nobel committee—and you may be surprised byhow simple they are. Reading them will give you a sense of how much weknew a long time ago, and also of how much we have learned in recentdecades.

(PDF) Thinking, fast and slow by daniel kahneman - PDFSLIDE.US (19)

Part 1

(PDF) Thinking, fast and slow by daniel kahneman - PDFSLIDE.US (20)

Two Systems

(PDF) Thinking, fast and slow by daniel kahneman - PDFSLIDE.US (21)

The Characters of the Story

To observe your mind in automatic mode, glance at the image below.

Figure 1

Your experience as you look at the woman’s face seamlessly combineswhat we normally call seeing and intuitive thinking. As surely and quickly asyou saw that the young woman’s hair is dark, you knew she is angry.Furthermore, what you saw extended into the future. You sensed that thiswoman is about to say some very unkind words, probably in a loud andstrident voice. A premonition of what she was going to do next came tomind automatically and effortlessly. You did not intend to assess her moodor to anticipate what she might do, and your reaction to the picture did nothave the feel of something you did. It just happened to you. It was aninstance of fast thinking.

Now look at the following problem:

17 × 24

You knew immediately that this is a multiplication problem, and probablyknew that you could solve it, with paper and pencil, if not without. You alsohad some vague intuitive knowledge of the range of possible results. Youwould be quick to recognize that both 12,609 and 123 are implausible.Without spending some time on the problem, however, you would not be

(PDF) Thinking, fast and slow by daniel kahneman - PDFSLIDE.US (22)

certain that the answer is not 568. A precise solution did not come to mind,and you felt that you could choose whether or not to engage in thecomputation. If you have not done so yet, you should attempt themultiplication problem now, completing at least part of it.

You experienced slow thinking as you proceeded through a sequence ofsteps. You first retrieved from memory the cognitive program formultiplication that you learned in school, then you implemented it. Carryingout the computation was a strain. You felt the burden of holding muchmaterial in memory, as you needed to keep track of where you were and ofwhere you were going, while holding on to the intermediate result. Theprocess was mental work: deliberate, effortful, and orderly—a prototype ofslow thinking. The computation was not only an event in your mind; yourbody was also involved. Your muscles tensed up, your blood pressurerose, and your heart rate increased. Someone looking closely at your eyeswhile you tackled this problem would have seen your pupils dilate. Yourpupils contracted back to normal size as soon as you ended your work—when you found the answer (which is 408, by the way) or when you gaveup.

Two Systems

Psychologists have been intensely interested for several decades in thetwo modagee fi Pn="cees of thinking evoked by the picture of the angrywoman and by the multiplication problem, and have offered many labels forthem. I adopt terms originally proposed by the psychologists KeithStanovich and Richard West, and will refer to two systems in the mind,System 1 and System 2.

System 1 operates automatically and quickly, with little or no effortand no sense of voluntary control.System 2 allocates attention to the effortful mental activities thatdemand it, including complex computations. The operations ofSystem 2 are often associated with the subjective experience ofagency, choice, and concentration.

The labels of System 1 and System 2 are widely used in psychology, but Igo further than most in this book, which you can read as a psychodramawith two characters.

When we think of ourselves, we identify with System 2, the conscious,

(PDF) Thinking, fast and slow by daniel kahneman - PDFSLIDE.US (23)

reasoning self that has beliefs, makes choices, and decides what to thinkabout and what to do. Although System 2 believes itself to be where theaction is, the automatic System 1 is the hero of the book. I describeSystem 1 as effortlessly originating impressions and feelings that are themain sources of the explicit beliefs and deliberate choices of System 2.The automatic operations of System 1 generate surprisingly complexpatterns of ideas, but only the slower System 2 can construct thoughts in anorderly series of steps. I also describe circumstances in which System 2takes over, overruling the freewheeling impulses and associations ofSystem 1. You will be invited to think of the two systems as agents withtheir individual abilities, limitations, and functions.

In rough order of complexity, here are some examples of the automaticactivities that are attributed to System 1:

Detect that one object is more distant than another.Orient to the source of a sudden sound.Complete the phrase “bread and…”Make a “disgust face” when shown a horrible picture.Detect hostility in a voice.Answer to 2 + 2 = ?Read words on large billboards.Drive a car on an empty road.Find a strong move in chess (if you are a chess master).Understand simple sentences.Recognize that a “meek and tidy soul with a passion for detail”resembles an occupational stereotype.

All these mental events belong with the angry woman—they occurautomatically and require little or no effort. The capabilities of System 1include innate skills that we share with other animals. We are bornprepared to perceive the world around us, recognize objects, orientattention, avoid losses, and fear spiders. Other mental activities becomefast and automatic through prolonged practice. System 1 has learnedassociations between ideas (the capital of France?); it has also learnedskills such as reading and understanding nuances of social situations.Some skills, such as finding strong chess moves, are acquired only byspecialized experts. Others are widely shared. Detecting the similarity of apersonality sketch to an occupatiohein occupatnal stereotype requiresbroad knowledge of the language and the culture, which most of us

(PDF) Thinking, fast and slow by daniel kahneman - PDFSLIDE.US (24)

possess. The knowledge is stored in memory and accessed withoutintention and without effort.

Several of the mental actions in the list are completely involuntary. Youcannot refrain from understanding simple sentences in your own languageor from orienting to a loud unexpected sound, nor can you prevent yourselffrom knowing that 2 + 2 = 4 or from thinking of Paris when the capital ofFrance is mentioned. Other activities, such as chewing, are susceptible tovoluntary control but normally run on automatic pilot. The control of attentionis shared by the two systems. Orienting to a loud sound is normally aninvoluntary operation of System 1, which immediately mobilizes thevoluntary attention of System 2. You may be able to resist turning towardthe source of a loud and offensive comment at a crowded party, but even ifyour head does not move, your attention is initially directed to it, at least fora while. However, attention can be moved away from an unwanted focus,primarily by focusing intently on another target.

The highly diverse operations of System 2 have one feature in common:they require attention and are disrupted when attention is drawn away.Here are some examples:

Brace for the starter gun in a race.Focus attention on the clowns in the circus.Focus on the voice of a particular person in a crowded and noisyroom.Look for a woman with white hair.Search memory to identify a surprising sound.Maintain a faster walking speed than is natural for you.Monitor the appropriateness of your behavior in a social situation.Count the occurrences of the letter a in a page of text.Tell someone your phone number.Park in a narrow space (for most people except garage attendants).Compare two washing machines for overall value.Fill out a tax form.Check the validity of a complex logical argument.

In all these situations you must pay attention, and you will perform less well,or not at all, if you are not ready or if your attention is directedinappropriately. System 2 has some ability to change the way System 1works, by programming the normally automatic functions of attention andmemory. When waiting for a relative at a busy train station, for example,

(PDF) Thinking, fast and slow by daniel kahneman - PDFSLIDE.US (25)

you can set yourself at will to look for a white-haired woman or a beardedman, and thereby increase the likelihood of detecting your relative from adistance. You can set your memory to search for capital cities that startwith N or for French existentialist novels. And when you rent a car atLondon’s Heathrow Airport, the attendant will probably remind you that “wedrive on the left side of the road over here.” In all these cases, you areasked to do something that does not come naturally, and you will find thatthe consistent maintenance of a set requires continuous exertion of at leastsome effort.

The often-used phrase “pay attention” is apt: you dispose of a limitedbudget of attention that you can allocate to activities, and if you try toi>Cyou try tgo beyond your budget, you will fail. It is the mark of effortfulactivities that they interfere with each other, which is why it is difficult orimpossible to conduct several at once. You could not compute the productof 17 × 24 while making a left turn into dense traffic, and you certainlyshould not try. You can do several things at once, but only if they are easyand undemanding. You are probably safe carrying on a conversation with apassenger while driving on an empty highway, and many parents havediscovered, perhaps with some guilt, that they can read a story to a childwhile thinking of something else.

Everyone has some awareness of the limited capacity of attention, andour social behavior makes allowances for these limitations. When thedriver of a car is overtaking a truck on a narrow road, for example, adultpassengers quite sensibly stop talking. They know that distracting thedriver is not a good idea, and they also suspect that he is temporarily deafand will not hear what they say.

Intense focusing on a task can make people effectively blind, even tostimuli that normally attract attention. The most dramatic demonstrationwas offered by Christopher Chabris and Daniel Simons in their book TheInvisible Gorilla. They constructed a short film of two teams passingbasketballs, one team wearing white shirts, the other wearing black. Theviewers of the film are instructed to count the number of passes made bythe white team, ignoring the black players. This task is difficult andcompletely absorbing. Halfway through the video, a woman wearing agorilla suit appears, crosses the court, thumps her chest, and moves on.The gorilla is in view for 9 seconds. Many thousands of people have seenthe video, and about half of them do not notice anything unusual. It is thecounting task—and especially the instruction to ignore one of the teams—that causes the blindness. No one who watches the video without that taskwould miss the gorilla. Seeing and orienting are automatic functions ofSystem 1, but they depend on the allocation of some attention to the

(PDF) Thinking, fast and slow by daniel kahneman - PDFSLIDE.US (26)

relevant stimulus. The authors note that the most remarkable observationof their study is that people find its results very surprising. Indeed, theviewers who fail to see the gorilla are initially sure that it was not there—they cannot imagine missing such a striking event. The gorilla studyillustrates two important facts about our minds: we can be blind to theobvious, and we are also blind to our blindness.

Plot Synopsis

The interaction of the two systems is a recurrent theme of the book, and abrief synopsis of the plot is in order. In the story I will tell, Systems 1 and 2are both active whenever we are awake. System 1 runs automatically andSystem 2 is normally in a comfortable low-effort mode, in which only afraction of its capacity is engaged. System 1 continuously generatessuggestions for System 2: impressions, intuitions, intentions, and feelings.If endorsed by System 2, impressions and intuitions turn into beliefs, andimpulses turn into voluntary actions. When all goes smoothly, which is mostof the time, System 2 adopts the suggestions of System 1 with little or nomodification. You generally believe your impressions and act on yourdesires, and that is fine—usually.

When System 1 runs into difficulty, it calls on System 2 to support moredetailed and specific processing that may solve the problem of themoment. System 2 is mobilized when a question arises for which System 1does not offer an answer, as probably happened to you when youencountered the multiplication problem 17 × 24. You can also feel a surgeof conscious attention whenever you are surprised. System 2 is activ">< 2is actated when an event is detected that violates the model of the worldthat System 1 maintains. In that world, lamps do not jump, cats do not bark,and gorillas do not cross basketball courts. The gorilla experimentdemonstrates that some attention is needed for the surprising stimulus tobe detected. Surprise then activates and orients your attention: you willstare, and you will search your memory for a story that makes sense of thesurprising event. System 2 is also credited with the continuous monitoringof your own behavior—the control that keeps you polite when you areangry, and alert when you are driving at night. System 2 is mobilized toincreased effort when it detects an error about to be made. Remember atime when you almost blurted out an offensive remark and note how hardyou worked to restore control. In summary, most of what you (your System2) think and do originates in your System 1, but System 2 takes over whenthings get difficult, and it normally has the last word.

The division of labor between System 1 and System 2 is highly efficient:

(PDF) Thinking, fast and slow by daniel kahneman - PDFSLIDE.US (27)

it minimizes effort and optimizes performance. The arrangement workswell most of the time because System 1 is generally very good at what itdoes: its models of familiar situations are accurate, its short-termpredictions are usually accurate as well, and its initial reactions tochallenges are swift and generally appropriate. System 1 has biases,however, systematic errors that it is prone to make in specifiedcircumstances. As we shall see, it sometimes answers easier questionsthan the one it was asked, and it has little understanding of logic andstatistics. One further limitation of System 1 is that it cannot be turned off. Ifyou are shown a word on the screen in a language you know, you will readit—unless your attention is totally focused elsewhere.

Conflict

Figure 2 is a variant of a classic experiment that produces a conflictbetween the two systems. You should try the exercise before reading on.

Figure 2

You were almost certainly successful in saying the correct words in bothtasks, and you surely discovered that some parts of each task were mucheasier than others. When you identified upper- and lowercase, the left-hand column was easy and the right-hand column caused you to slow down

(PDF) Thinking, fast and slow by daniel kahneman - PDFSLIDE.US (28)

hand column was easy and the right-hand column caused you to slow downand perhaps to stammer or stumble. When you named the position ofwords, the left-hand column was difficult and the right-hand column wasmuch easier.

These tasks engage System 2, because saying “upper/lower” or“right/left” is not what you routinely do when looking down a column ofwords. One of the things you did to set yourself for the task was to programyour memory so that the relevant words (upper and lower for the first task)were “on the tip of your tongue.” The prioritizing of the chosen words iseffective and the mild temptation to read other words was fairly easy toresist when you went through the first column. But the second column wasdifferent, because it contained words for which you were set, and you couldnot ignore them. You were mostly able to respond correctly, butovercoming the competing response was a strain, and it slowed you down.You experienced a conflict between a task that you intended to carry outand an automatic response that interfered with it.

Conflict between an automatic reaction and an intention to conWhetionto ctrol it is common in our lives. We are all familiar with the experience oftrying not to stare at the oddly dressed couple at the neighboring table in arestaurant. We also know what it is like to force our attention on a boringbook, when we constantly find ourselves returning to the point at which thereading lost its meaning. Where winters are hard, many drivers havememories of their car skidding out of control on the ice and of the struggleto follow well-rehearsed instructions that negate what they would naturallydo: “Steer into the skid, and whatever you do, do not touch the brakes!”And every human being has had the experience of not telling someone togo to hell. One of the tasks of System 2 is to overcome the impulses ofSystem 1. In other words, System 2 is in charge of self-control.

Illusions

To appreciate the autonomy of System 1, as well as the distinctionbetween impressions and beliefs, take a good look at figure 3.

This picture is unremarkable: two horizontal lines of different lengths,with fins appended, pointing in different directions. The bottom line isobviously longer than the one above it. That is what we all see, and wenaturally believe what we see. If you have already encountered this image,however, you recognize it as the famous Müller-Lyer illusion. As you caneasily confirm by measuring them with a ruler, the horizontal lines are infact identical in length.

(PDF) Thinking, fast and slow by daniel kahneman - PDFSLIDE.US (29)

Figure 3

Now that you have measured the lines, you—your System 2, theconscious being you call “I”—have a new belief: you know that the lines areequally long. If asked about their length, you will say what you know. But yousti ll see the bottom line as longer. You have chosen to believe themeasurement, but you cannot prevent System 1 from doing its thing; youcannot decide to see the lines as equal, although you know they are. Toresist the illusion, there is only one thing you can do: you must learn tomistrust your impressions of the length of lines when fins are attached tothem. To implement that rule, you must be able to recognize the illusorypattern and recall what you know about it. If you can do this, you will neveragain be fooled by the Müller-Lyer illusion. But you will still see one line aslonger than the other.

Not all illusions are visual. There are illusions of thought, which we callcognitive illusions. As a graduate student, I attended some courses on theart and science of psychotherapy. During one of these lectures, ourteacher imparted a morsel of clinical wisdom. This is what he told us: “Youwill from time to time meet a patient who shares a disturbing tale ofmultiple mistakes in his previous treatment. He has been seen by severalclinicians, and all failed him. The patient can lucidly describe how histherapists misunderstood him, but he has quickly perceived that you aredifferent. You share the same feeling, are convinced that you understandhim, and will be able to help.” At this point my teacher raised his voice ashe said, “Do not even think of taking on this patient! Throw him out of theoffice! He is most likely a psychopath and you will not be able to help him.”

Many years later I learned that the teacher had warned us againstpsychopathic charm, and the leading authority in the strn y in the udy of

(PDF) Thinking, fast and slow by daniel kahneman - PDFSLIDE.US (30)

psychopathy confirmed that the teacher’s advice was sound. The analogyto the Müller-Lyer illusion is close. What we were being taught was not howto feel about that patient. Our teacher took it for granted that the sympathywe would feel for the patient would not be under our control; it would arisefrom System 1. Furthermore, we were not being taught to be generallysuspicious of our feelings about patients. We were told that a strongattraction to a patient with a repeated history of failed treatment is adanger sign—like the fins on the parallel lines. It is an illusion—a cognitiveillusion—and I (System 2) was taught how to recognize it and advised notto believe it or act on it.

The question that is most often asked about cognitive illusions iswhether they can be overcome. The message of these examples is notencouraging. Because System 1 operates automatically and cannot beturned off at will, errors of intuitive thought are often difficult to prevent.Biases cannot always be avoided, because System 2 may have no clue tothe error. Even when cues to likely errors are available, errors can beprevented only by the enhanced monitoring and effortful activity of System2. As a way to live your life, however, continuous vigilance is notnecessarily good, and it is certainly impractical. Constantly questioning ourown thinking would be impossibly tedious, and System 2 is much too slowand inefficient to serve as a substitute for System 1 in making routinedecisions. The best we can do is a compromise: learn to recognizesituations in which mistakes are likely and try harder to avoid significantmistakes when the stakes are high. The premise of this book is that it iseasier to recognize other people’s mistakes than our own.

Useful Fictions

You have been invited to think of the two systems as agents within themind, with their individual personalities, abilities, and limitations. I will oftenuse sentences in which the systems are the subjects, such as, “System 2calculates products.”

The use of such language is considered a sin in the professional circlesin which I travel, because it seems to explain the thoughts and actions of aperson by the thoughts and actions of little people inside the person’shead. Grammatically the sentence about System 2 is similar to “The butlersteals the petty cash.” My colleagues would point out that the butler’s actionactually explains the disappearance of the cash, and they rightly questionwhether the sentence about System 2 explains how products arecalculated. My answer is that the brief active sentence that attributescalculation to System 2 is intended as a description, not an explanation. It

(PDF) Thinking, fast and slow by daniel kahneman - PDFSLIDE.US (31)

is meaningful only because of what you already know about System 2. It isshorthand for the following: “Mental arithmetic is a voluntary activity thatrequires effort, should not be performed while making a left turn, and isassociated with dilated pupils and an accelerated heart rate.”

Similarly, the statement that “highway driving under routine conditions isleft to System 1” means that steering the car around a bend is automaticand almost effortless. It also implies that an experienced driver can driveon an empty highway while conducting a conversation. Finally, “System 2prevented James from reacting foolishly to the insult” means that Jameswould have been more aggressive in his response if his capacity foreffortful control had been disrupted (for example, if he had been drunk).

System 1 and System 2 are so central to the story I tell in this book that Imust make it absolutely clear that they are217at they a fictitiouscharacters. Systems 1 and 2 are not systems in the standard sense ofentities with interacting aspects or parts. And there is no one part of thebrain that either of the systems would call home. You may well ask: What isthe point of introducing fictitious characters with ugly names into a seriousbook? The answer is that the characters are useful because of somequirks of our minds, yours and mine. A sentence is understood more easilyif it describes what an agent (System 2) does than if it describes whatsomething is, what properties it has. In other words, “System 2” is a bettersubject for a sentence than “mental arithmetic.” The mind—especiallySystem 1—appears to have a special aptitude for the construction andinterpretation of stories about active agents, who have personalities,habits, and abilities. You quickly formed a bad opinion of the thievingbutler, you expect more bad behavior from him, and you will remember himfor a while. This is also my hope for the language of systems.

Why call them System 1 and System 2 rather than the more descriptive“automatic system” and “effortful system”? The reason is simple:“Automatic system” takes longer to say than “System 1” and thereforetakes more space in your working memory. This matters, becauseanything that occupies your working memory reduces your ability to think.You should treat “System 1” and “System 2” as nicknames, like Bob andJoe, identifying characters that you will get to know over the course of thisbook. The fictitious systems make it easier for me to think about judgmentand choice, and will make it easier for you to understand what I say.

Speaking of System 1 and System 2

(PDF) Thinking, fast and slow by daniel kahneman - PDFSLIDE.US (32)

“He had an impression, but some of his impressions areillusions.”

“This was a pure System 1 response. She reacted to the threatbefore she recognized it.”

“This is your System 1 talking. Slow down and let your System 2take control.”

(PDF) Thinking, fast and slow by daniel kahneman - PDFSLIDE.US (33)

Attention and Effort

In the unlikely event of this book being made into a film, System 2 would bea supporting character who believes herself to be the hero. The definingfeature of System 2, in this story, is that its operations are effortful, and oneof its main characteristics is laziness, a reluctance to invest more effortthan is strictly necessary. As a consequence, the thoughts and actions thatSystem 2 believes it has chosen are often guided by the figure at thecenter of the story, System 1. However, there are vital tasks that onlySystem 2 can perform because they require effort and acts of self-controlin which the intuitions and impulses of System 1 are overcome.

Mental Effort

If you wish to experience your System 2 working at full tilt, the followingexercise will do; it should br"0%e ca Tting you to the limits of your cognitiveabilities within 5 seconds. To start, make up several strings of 4 digits, alldifferent, and write each string on an index card. Place a blank card on topof the deck. The task that you will perform is called Add-1. Here is how itgoes:

Start beating a steady rhythm (or better yet, set a metronome at1/sec). Remove the blank card and read the four digits aloud.Wait for two beats, then report a string in which each of theoriginal digits is incremented by 1. If the digits on the card are5294, the correct response is 6305. Keeping the rhythm isimportant.

Few people can cope with more than four digits in the Add-1 task, but ifyou want a harder challenge, please try Add-3.

If you would like to know what your body is doing while your mind is hardat work, set up two piles of books on a sturdy table, place a video cameraon one and lean your chin on the other, get the video going, and stare atthe camera lens while you work on Add-1 or Add-3 exercises. Later, youwill find in the changing size of your pupils a faithful record of how hard youworked.

I have a long personal history with the Add-1 task. Early in my career Ispent a year at the University of Michigan, as a visitor in a laboratory thatstudied hypnosis. Casting about for a useful topic of research, I found anarticle in Scientific American in which the psychologist Eckhard Hessdescribed the pupil of the eye as a window to the soul. I reread it recently

(PDF) Thinking, fast and slow by daniel kahneman - PDFSLIDE.US (34)

and again found it inspiring. It begins with Hess reporting that his wife hadnoticed his pupils widening as he watched beautiful nature pictures, and itends with two striking pictures of the same good-looking woman, whosomehow appears much more attractive in one than in the other. There isonly one difference: the pupils of the eyes appear dilated in the attractivepicture and constricted in the other. Hess also wrote of belladonna, a pupil-dilating substance that was used as a cosmetic, and of bazaar shopperswho wear dark glasses in order to hide their level of interest frommerchants.

One of Hess’s findings especially captured my attention. He had noticedthat the pupils are sensitive indicators of mental effort—they dilatesubstantially when people multiply two-digit numbers, and they dilate moreif the problems are hard than if they are easy. His observations indicatedthat the response to mental effort is distinct from emotional arousal. Hess’swork did not have much to do with hypnosis, but I concluded that the ideaof a visible indication of mental effort had promise as a research topic. Agraduate student in the lab, Jackson Beatty, shared my enthusiasm and wegot to work.

Beatty and I developed a setup similar to an optician’s examinationroom, in which the experimental participant leaned her head on a chin-and-forehead rest and stared at a camera while listening to prerecordedinformation and answering questions on the recorded beats of ametronome. The beats triggered an infrared flash every second, causing apicture to be taken. At the end of each experimental session, we wouldrush to have the film developed, project the images of the pupil on ascreen, and go to work with a ruler. The method was a perfect fit for youngand impatient researchers: we knew our results almost immediately, andthey always told a clear story.

Beatty and I focused on paced tasks, such as Add-1, in which we knewprecisely what was on the subject’s mind at any time. We recorded stringsof digits on beats of the metronome and instructed the subject to repeat ortransform the digits one indigits onby one, maintaining the same rhythm.We soon discovered that the size of the pupil varied second by second,reflecting the changing demands of the task. The shape of the responsewas an inverted V. As you experienced it if you tried Add-1 or Add-3, effortbuilds up with every added digit that you hear, reaches an almostintolerable peak as you rush to produce a transformed string during andimmediately after the pause, and relaxes gradually as you “unload” yourshort-term memory. The pupil data corresponded precisely to subjectiveexperience: longer strings reliably caused larger dilations, thetransformation task compounded the effort, and the peak of pupil sizecoincided with maximum effort. Add-1 with four digits caused a larger

(PDF) Thinking, fast and slow by daniel kahneman - PDFSLIDE.US (35)

dilation than the task of holding seven digits for immediate recall. Add-3,which is much more difficult, is the most demanding that I ever observed. Inthe first 5 seconds, the pupil dilates by about 50% of its original area andheart rate increases by about 7 beats per minute. This is as hard aspeople can work—they give up if more is asked of them. When weexposed our subjects to more digits than they could remember, their pupilsstopped dilating or actually shrank.

We worked for some months in a spacious basement suite in which wehad set up a closed-circuit system that projected an image of the subject’spupil on a screen in the corridor; we also could hear what was happeningin the laboratory. The diameter of the projected pupil was about a foot;watching it dilate and contract when the participant was at work was afascinating sight, quite an attraction for visitors in our lab. We amusedourselves and impressed our guests by our ability to divine when theparticipant gave up on a task. During a mental multiplication, the pupilnormally dilated to a large size within a few seconds and stayed large aslong as the individual kept working on the problem; it contractedimmediately when she found a solution or gave up. As we watched fromthe corridor, we would sometimes surprise both the owner of the pupil andour guests by asking, “Why did you stop working just now?” The answerfrom inside the lab was often, “How did you know?” to which we wouldreply, “We have a window to your soul.”

The casual observations we made from the corridor were sometimes asinformative as the formal experiments. I made a significant discovery as Iwas idly watching a woman’s pupil during a break between two tasks. Shehad kept her position on the chin rest, so I could see the image of her eyewhile she engaged in routine conversation with the experimenter. I wassurprised to see that the pupil remained small and did not noticeably dilateas she talked and listened. Unlike the tasks that we were studying, themundane conversation apparently demanded little or no effort—no morethan retaining two or three digits. This was a eureka moment: I realized thatthe tasks we had chosen for study were exceptionally effortful. An imagecame to mind: mental life—today I would speak of the life of System 2—isnormally conducted at the pace of a comfortable walk, sometimesinterrupted by episodes of jogging and on rare occasions by a franticsprint. The Add-1 and Add-3 exercises are sprints, and casual chatting isa stroll.

We found that people, when engaged in a mental sprint, may becomeeffectively blind. The authors of The Invisible Gorilla had made the gorilla“invisible” by keeping the observers intensely busy counting passes. Wereported a rather less dramatic example of blindness during Add-1. Our

(PDF) Thinking, fast and slow by daniel kahneman - PDFSLIDE.US (36)

subjects were exposed to a series of rapidly flashing letters while theyworked. They were told to give the task complete priority, but they werealso asked to report, at the end of the digit task, whether the letter K hadappeared at any rored at antime during the trial. The main finding was thatthe ability to detect and report the target letter changed in the course of the10 seconds of the exercise. The observers almost never missed a K thatwas shown at the beginning or near the end of the Add-1 task but theymissed the target almost half the time when mental effort was at its peak,although we had pictures of their wide-open eye staring straight at it.Failures of detection followed the same inverted-V pattern as the dilatingpupil. The similarity was reassuring: the pupil was a good measure of thephysical arousal that accompanies mental effort, and we could go aheadand use it to understand how the mind works.

Much like the electricity meter outside your house or apartment, thepupils offer an index of the current rate at which mental energy is used. Theanalogy goes deep. Your use of electricity depends on what you choose todo, whether to light a room or toast a piece of bread. When you turn on abulb or a toaster, it draws the energy it needs but no more. Similarly, wedecide what to do, but we have limited control over the effort of doing it.Suppose you are shown four digits, say, 9462, and told that your lifedepends on holding them in memory for 10 seconds. However much youwant to live, you cannot exert as much effort in this task as you would beforced to invest to complete an Add-3 transformation on the same digits.

System 2 and the electrical circuits in your home both have limitedcapacity, but they respond differently to threatened overload. A breakertrips when the demand for current is excessive, causing all devices on thatcircuit to lose power at once. In contrast, the response to mental overloadis selective and precise: System 2 protects the most important activity, soit receives the attention it needs; “spare capacity” is allocated second bysecond to other tasks. In our version of the gorilla experiment, weinstructed the participants to assign priority to the digit task. We know thatthey followed that instruction, because the timing of the visual target had noeffect on the main task. If the critical letter was presented at a time of highdemand, the subjects simply did not see it. When the transformation taskwas less demanding, detection performance was better.

The sophisticated allocation of attention has been honed by a longevolutionary history. Orienting and responding quickly to the gravest threatsor most promising opportunities improved the chance of survival, and thiscapability is certainly not restricted to humans. Even in modern humans,System 1 takes over in emergencies and assigns total priority to self-protective actions. Imagine yourself at the wheel of a car that unexpectedly

(PDF) Thinking, fast and slow by daniel kahneman - PDFSLIDE.US (37)

skids on a large oil slick. You will find that you have responded to the threatbefore you became fully conscious of it.

Beatty and I worked together for only a year, but our collaboration had alarge effect on our subsequent careers. He eventually became the leadingauthority on “cognitive pupillometry,” and I wrote a book titled Attention andEffort, which was based in large part on what we learned together and onfollow-up research I did at Harvard the following year. We learned a greatdeal about the working mind—which I now think of as System 2—frommeasuring pupils in a wide variety of tasks.

As you become skilled in a task, its demand for energy diminishes.Studies of the brain have shown that the pattern of activity associated withan action changes as skill increases, with fewer brain regions involved.Talent has similar effects. Highly intelligent individuals need less effort tosolve the same problems, as indicated by both pupil size and brain activity.A general “law of least effort” appd t” alies to cognitive as well as physicalexertion. The law asserts that if there are several ways of achieving thesame goal, people will eventually gravitate to the least demanding courseof action. In the economy of action, effort is a cost, and the acquisition ofskill is driven by the balance of benefits and costs. Laziness is built deepinto our nature.

The tasks that we studied varied considerably in their effects on thepupil. At baseline, our subjects were awake, aware, and ready to engagein a task—probably at a higher level of arousal and cognitive readinessthan usual. Holding one or two digits in memory or learning to associate aword with a digit (3 = door) produced reliable effects on momentaryarousal above that baseline, but the effects were minuscule, only 5% of theincrease in pupil diameter associated with Add-3. A task that requireddiscriminating between the pitch of two tones yielded significantly largerdilations. Recent research has shown that inhibiting the tendency to readdistracting words (as in figure 2 of the preceding chapter) also inducesmoderate effort. Tests of short-term memory for six or seven digits weremore effortful. As you can experience, the request to retrieve and say aloudyour phone number or your spouse’s birthday also requires a brief butsignificant effort, because the entire string must be held in memory as aresponse is organized. Mental multiplication of two-digit numbers and theAdd-3 task are near the limit of what most people can do.

What makes some cognitive operations more demanding and effortfulthan others? What outcomes must we purchase in the currency ofattention? What can System 2 do that System 1 cannot? We now havetentative answers to these questions.

Effort is required to maintain simultaneously in memory several ideas

(PDF) Thinking, fast and slow by daniel kahneman - PDFSLIDE.US (38)

that require separate actions, or that need to be combined according to arule—rehearsing your shopping list as you enter the supermarket,choosing between the fish and the veal at a restaurant, or combining asurprising result from a survey with the information that the sample wassmall, for example. System 2 is the only one that can follow rules, compareobjects on several attributes, and make deliberate choices betweenoptions. The automatic System 1 does not have these capabilities. System1 detects simple relations (“they are all alike,” “the son is much taller thanthe father”) and excels at integrating information about one thing, but itdoes not deal with multiple distinct topics at once, nor is it adept at usingpurely statistical information. System 1 will detect that a person describedas “a meek and tidy soul, with a need for order and structure, and apassion for detail” resembles a caricature librarian, but combining thisintuition with knowledge about the small number of librarians is a task thatonly System 2 can perform—if System 2 knows how to do so, which is trueof few people.

A crucial capability of System 2 is the adoption of “task sets”: it canprogram memory to obey an instruction that overrides habitual responses.Consider the following: Count all occurrences of the letter f in this page.This is not a task you have ever performed before and it will not comenaturally to you, but your System 2 can take it on. It will be effortful to setyourself up for this exercise, and effortful to carry it out, though you willsurely improve with practice. Psychologists speak of “executive control” todescribe the adoption and termination of task sets, and neuroscientistshave identified the main regions of the brain that serve the executivefunction. One of these regions is involved whenever a conflict must beresolved. Another is the prefrontal area of the brain, a region that issubstantially more developed in humans tht un humans an in otherprimates, and is involved in operations that we associate with intelligence.

Now suppose that at the end of the page you get another instruction:count all the commas in the next page. This will be harder, because you willhave to overcome the newly acquired tendency to focus attention on theletter f. One of the significant discoveries of cognitive psychologists inrecent decades is that switching from one task to another is effortful,especially under time pressure. The need for rapid switching is one of thereasons that Add-3 and mental multiplication are so difficult. To performthe Add-3 task, you must hold several digits in your working memory at thesame time, associating each with a particular operation: some digits are inthe queue to be transformed, one is in the process of transformation, andothers, already transformed, are retained for reporting. Modern tests ofworking memory require the individual to switch repeatedly between two

(PDF) Thinking, fast and slow by daniel kahneman - PDFSLIDE.US (39)

demanding tasks, retaining the results of one operation while performingthe other. People who do well on these tests tend to do well on tests ofgeneral intelligence. However, the ability to control attention is not simply ameasure of intelligence; measures of efficiency in the control of attentionpredict performance of air traffic controllers and of Israeli Air Force pilotsbeyond the effects of intelligence.

Time pressure is another driver of effort. As you carried out the Add-3exercise, the rush was imposed in part by the metronome and in part bythe load on memory. Like a juggler with several balls in the air, you cannotafford to slow down; the rate at which material decays in memory forcesthe pace, driving you to refresh and rehearse information before it is lost.Any task that requires you to keep several ideas in mind at the same timehas the same hurried character. Unless you have the good fortune of acapacious working memory, you may be forced to work uncomfortablyhard. The most effortful forms of slow thinking are those that require you tothink fast.

You surely observed as you performed Add-3 how unusual it is for yourmind to work so hard. Even if you think for a living, few of the mental tasksin which you engage in the course of a working day are as demanding asAdd-3, or even as demanding as storing six digits for immediate recall.We normally avoid mental overload by dividing our tasks into multiple easysteps, committing intermediate results to long-term memory or to paperrather than to an easily overloaded working memory. We cover longdistances by taking our time and conduct our mental lives by the law ofleast effort.

Speaking of Attention and Effort

“I won’t try to solve this while driving. This is a pupil-dilating task. Itrequires mental effort!”

“The law of least effort is operating here. He will think as little aspossible.”

“She did not forget about the meeting. She was completelyfocused on something else when the meeting was set and shejust didn’t hear you.”

(PDF) Thinking, fast and slow by daniel kahneman - PDFSLIDE.US (40)

“What came quickly to my mind was an intuition from System 1. I’llhave to start over and search my memory deliberately.”

(PDF) Thinking, fast and slow by daniel kahneman - PDFSLIDE.US (41)

The Lazy Controller

I spend a few months each year in Berkeley, and one of my greatpleasures there is a daily four-mile walk on a marked path in the hills, witha fine view of San Francisco Bay. I usually keep track of my time and havelearned a fair amount about effort from doing so. I have found a speed,about 17 minutes for a mile, which I experience as a stroll. I certainly exertphysical effort and burn more calories at that speed than if I sat in arecliner, but I experience no strain, no conflict, and no need to push myself.I am also able to think and work while walking at that rate. Indeed, I suspectthat the mild physical arousal of the walk may spill over into greater mentalalertness.

System 2 also has a natural speed. You expend some mental energy inrandom thoughts and in monitoring what goes on around you even whenyour mind does nothing in particular, but there is little strain. Unless you arein a situation that makes you unusually wary or self-conscious, monitoringwhat happens in the environment or inside your head demands little effort.You make many small decisions as you drive your car, absorb someinformation as you read the newspaper, and conduct routine exchanges ofpleasantries with a spouse or a colleague, all with little effort and no strain.Just like a stroll.

It is normally easy and actually quite pleasant to walk and think at thesame time, but at the extremes these activities appear to compete for thelimited resources of System 2. You can confirm this claim by a simpleexperiment. While walking comfortably with a friend, ask him to compute23 × 78 in his head, and to do so immediately. He will almost certainly stopin his tracks. My experience is that I can think while strolling but cannotengage in mental work that imposes a heavy load on short-term memory. IfI must construct an intricate argument under time pressure, I would ratherbe still, and I would prefer sitting to standing. Of course, not all slowthinking requires that form of intense concentration and effortfulcomputation—I did the best thinking of my life on leisurely walks withAmos.

Accelerating beyond my strolling speed completely changes theexperience of walking, because the transition to a faster walk brings abouta sharp deterioration in my ability to think coherently. As I speed up, myattention is drawn with increasing frequency to the experience of walkingand to the deliberate maintenance of the faster pace. My ability to bring atrain of thought to a conclusion is impaired accordingly. At the highestspeed I can sustain on the hills, about 14 minutes for a mile, I do not eventry to think of anything else. In addition to the physical effort of moving my

(PDF) Thinking, fast and slow by daniel kahneman - PDFSLIDE.US (42)

body rapidly along the path, a mental effort of self-control is needed toresist the urge to slow down. Self-control and deliberate thought apparentlydraw on the same limited budget of effort.

For most of us, most of the time, the maintenance of a coherent train ofthought and the occasional engagement in effortful thinking also requireself-control. Although I have not conducted a systematic survey, I suspectthat frequent switching of tasks and speeded-up mental work are notintrinsically pleasurable, and that people avoid them when possible. This ishow the law of least effort comes to be a law. Even in the absence of timepressure, maintaining a coherent train of thought requires discipline. Anobserver of the number of times I look at e-mail or investigate therefrigerator during an hour of writing could wahene dd reasonably infer anurge to escape and conclude that keeping at it requires more self-controlthan I can readily muster.

Fortunately, cognitive work is not always aversive, and peoplesometimes expend considerable effort for long periods of time withouthaving to exert willpower. The psychologist Mihaly Csikszentmihalyi(pronounced six-cent-mihaly) has done more than anyone else to study thisstate of effortless attending, and the name he proposed for it, flow, hasbecome part of the language. People who experience flow describe it as“a state of effortless concentration so deep that they lose their sense oftime, of themselves, of their problems,” and their descriptions of the joy ofthat state are so compelling that Csikszentmihalyi has called it an “optimalexperience.” Many activities can induce a sense of flow, from painting toracing motorcycles—and for some fortunate authors I know, even writing abook is often an optimal experience. Flow neatly separates the two formsof effort: concentration on the task and the deliberate control of attention.Riding a motorcycle at 150 miles an hour and playing a competitive gameof chess are certainly very effortful. In a state of flow, however, maintainingfocused attention on these absorbing activities requires no exertion of self-control, thereby freeing resources to be directed to the task at hand.

The Busy and Depleted System 2

It is now a well-established proposition that both self-control and cognitiveeffort are forms of mental work. Several psychological studies have shownthat people who are simultaneously challenged by a demanding cognitivetask and by a temptation are more likely to yield to the temptation. Imaginethat you are asked to retain a list of seven digits for a minute or two. Youare told that remembering the digits is your top priority. While yourattention is focused on the digits, you are offered a choice between two

(PDF) Thinking, fast and slow by daniel kahneman - PDFSLIDE.US (43)

desserts: a sinful chocolate cake and a virtuous fruit salad. The evidencesuggests that you would be more likely to select the tempting chocolatecake when your mind is loaded with digits. System 1 has more influenceon behavior when System 2 is busy, and it has a sweet tooth.

People who are cognitively busy are also more likely to make selfishchoices, use sexist language, and make superficial judgments in socialsituations. Memorizing and repeating digits loosens the hold of System 2on behavior, but of course cognitive load is not the only cause ofweakened self-control. A few drinks have the same effect, as does asleepless night. The self-control of morning people is impaired at night; thereverse is true of night people. Too much concern about how well one isdoing in a task sometimes disrupts performance by loading short-termmemory with pointless anxious thoughts. The conclusion is straightforward:self-control requires attention and effort. Another way of saying this is thatcontrolling thoughts and behaviors is one of the tasks that System 2performs.

A series of surprising experiments by the psychologist Roy Baumeisterand his colleagues has shown conclusively that all variants of voluntaryeffort—cognitive, emotional, or physical—draw at least partly on a sharedpool of mental energy. Their experiments involve successive rather thansimultaneous tasks.

Baumeister’s group has repeatedly found that an effort of will or self-control is tiring; if you have had to force yourself to do something, you areless willing or less able to exert self-control when the next challenge comesaround. The phenomenon has been named ego depletion. In a typicaldemo thypical denstration, participants who are instructed to stifle theiremotional reaction to an emotionally charged film will later perform poorlyon a test of physical stamina—how long they can maintain a strong grip ona dynamometer in spite of increasing discomfort. The emotional effort inthe first phase of the experiment reduces the ability to withstand the pain ofsustained muscle contraction, and ego-depleted people thereforesuccumb more quickly to the urge to quit. In another experiment, peopleare first depleted by a task in which they eat virtuous foods such asradishes and celery while resisting the temptation to indulge in chocolateand rich cookies. Later, these people will give up earlier than normal whenfaced with a difficult cognitive task.

The list of situations and tasks that are now known to deplete self-controlis long and varied. All involve conflict and the need to suppress a naturaltendency. They include:

avoiding the thought of white bearsinhibiting the emotional response to a stirring film

(PDF) Thinking, fast and slow by daniel kahneman - PDFSLIDE.US (44)

making a series of choices that involve conflicttrying to impress othersresponding kindly to a partner’s bad behaviorinteracting with a person of a different race (for prejudicedindividuals)

The list of indications of depletion is also highly diverse:

deviating from one’s dietoverspending on impulsive purchasesreacting aggressively to provocationpersisting less time in a handgrip taskperforming poorly in cognitive tasks and logical decision making

The evidence is persuasive: activities that impose high demands onSystem 2 require self-control, and the exertion of self-control is depletingand unpleasant. Unlike cognitive load, ego depletion is at least in part aloss of motivation. After exerting self-control in one task, you do not feellike making an effort in another, although you could do it if you really had to.In several experiments, people were able to resist the effects of egodepletion when given a strong incentive to do so. In contrast, increasingeffort is not an option when you must keep six digits in short-term memorywhile performing a task. Ego depletion is not the same mental state ascognitive busyness.

The most surprising discovery made by Baumeister’s group shows, ashe puts it, that the idea of mental energy is more than a mere metaphor.The nervous system consumes more glucose than most other parts of thebody, and effortful mental activity appears to be especially expensive in thecurrency of glucose. When you are actively involved in difficult cognitivereasoning or engaged in a task that requires self-control, your bloodglucose level drops. The effect is analogous to a runner who draws downglucose stored in her muscles during a sprint. The bold implication of thisidea is that the effects of ego depletion could be undone by ingestingglucose, and Baumeister and his colleagues have confirmed thishypothesis n ohypothesiin several experiments.

Volunteers in one of their studies watched a short silent film of a womanbeing interviewed and were asked to interpret her body language. Whilethey were performing the task, a series of words crossed the screen inslow succession. The participants were specifically instructed to ignore thewords, and if they found their attention drawn away they had to refocus theirconcentration on the woman’s behavior. This act of self-control was knownto cause ego depletion. All the volunteers drank some lemonade before

(PDF) Thinking, fast and slow by daniel kahneman - PDFSLIDE.US (45)

participating in a second task. The lemonade was sweetened with glucosefor half of them and with Splenda for the others. Then all participants weregiven a task in which they needed to overcome an intuitive response to getthe correct answer. Intuitive errors are normally much more frequent amongego-depleted people, and the drinkers of Splenda showed the expecteddepletion effect. On the other hand, the glucose drinkers were notdepleted. Restoring the level of available sugar in the brain had preventedthe deterioration of performance. It will take some time and much furtherresearch to establish whether the tasks that cause glucose-depletion alsocause the momentary arousal that is reflected in increases of pupil sizeand heart rate.

A disturbing demonstration of depletion effects in judgment was recentlyreported in the Proceedings of the National Academy of Sciences. Theunwitting participants in the study were eight parole judges in Israel. Theyspend entire days reviewing applications for parole. The cases arepresented in random order, and the judges spend little time on each one,an average of 6 minutes. (The default decision is denial of parole; only35% of requests are approved. The exact time of each decision isrecorded, and the times of the judges’ three food breaks—morning break,lunch, and afternoon break—during the day are recorded as well.) Theauthors of the study plotted the proportion of approved requests againstthe time since the last food break. The proportion spikes after each meal,when about 65% of requests are granted. During the two hours or so untilthe judges’ next feeding, the approval rate drops steadily, to about zero justbefore the meal. As you might expect, this is an unwelcome result and theauthors carefully checked many alternative explanations. The best possibleaccount of the data provides bad news: tired and hungry judges tend to fallback on the easier default position of denying requests for parole. Bothfatigue and hunger probably play a role.

The Lazy System 2

One of the main functions of System 2 is to monitor and control thoughtsand actions “suggested” by System 1, allowing some to be expresseddirectly in behavior and suppressing or modifying others.

For an example, here is a simple puzzle. Do not try to solve it but listento your intuition:

A bat and ball cost $1.10.The bat costs one dollar more than the ball.How much does the ball cost?

(PDF) Thinking, fast and slow by daniel kahneman - PDFSLIDE.US (46)

A number came to your mind. The number, of course, is 10: 10¢. Thedistinctive mark of this easy puzzle is that it evokes an answer that isintuitive, appealing, and wrong. Do the math, and you will see. If the ballcosts 10¢, then the total cost will be $1.20 (10¢ for the ball and $1.10 forthe bat), not $1.10. The correct answer is 5¢. It%">5¢. is safe to assumethat the intuitive answer also came to the mind of those who ended up withthe correct number—they somehow managed to resist the intuition.

Shane Frederick and I worked together on a theory of judgment basedon two systems, and he used the bat-and-ball puzzle to study a centralquestion: How closely does System 2 monitor the suggestions of System1? His reasoning was that we know a significant fact about anyone whosays that the ball costs 10¢: that person did not actively check whether theanswer was correct, and her System 2 endorsed an intuitive answer that itcould have rejected with a small investment of effort. Furthermore, we alsoknow that the people who give the intuitive answer have missed an obvioussocial cue; they should have wondered why anyone would include in aquestionnaire a puzzle with such an obvious answer. A failure to check isremarkable because the cost of checking is so low: a few seconds ofmental work (the problem is moderately difficult), with slightly tensedmuscles and dilated pupils, could avoid an embarrassing mistake. Peoplewho say 10¢ appear to be ardent followers of the law of least effort. Peoplewho avoid that answer appear to have more active minds.

Many thousands of university students have answered the bat-and-ballpuzzle, and the results are shocking. More than 50% of students atHarvard, MIT, and Princeton ton gave the intuitive—incorrect—answer. Atless selective universities, the rate of demonstrable failure to check was inexcess of 80%. The bat-and-ball problem is our first encounter with anobservation that will be a recurrent theme of this book: many people areoverconfident, prone to place too much faith in their intuitions. Theyapparently find cognitive effort at least mildly unpleasant and avoid it asmuch as possible.

Now I will show you a logical argument—two premises and a conclusion.Try to determine, as quickly as you can, if the argument is logically valid.Does the conclusion follow from the premises?

All roses are flowers.Some flowers fade quickly.Therefore some roses fade quickly.

A large majority of college students endorse this syllogism as valid. In factthe argument is flawed, because it is possible that there are no rosesamong the flowers that fade quickly. Just as in the bat-and-ball problem, a

(PDF) Thinking, fast and slow by daniel kahneman - PDFSLIDE.US (47)

plausible answer comes to mind immediately. Overriding it requires hardwork—the insistent idea that “it’s true, it’s true!” makes it difficult to checkthe logic, and most people do not take the trouble to think through theproblem.

This experiment has discouraging implications for reasoning in everydaylife. It suggests that when people believe a conclusion is true, they are alsovery likely to believe arguments that appear to support it, even when thesearguments are unsound. If System 1 is involved, the conclusion comes firstand the arguments follow.

Next, consider the following question and answer it quickly beforereading on:

How many murders occur in the state of Michigan in one year?

The question, which was also devised by Shane Frederick, is again achallenge to System 2. The “trick” is whether the respondent will rememberthat Detroit, a high-crime c thigh-crimeity, is in Michigan. College studentsin the United States know this fact and will correctly identify Detroit as thelargest city in Michigan. But knowledge of a fact is not all-or-none. Factsthat we know do not always come to mind when we need them. Peoplewho remember that Detroit is in Michigan give higher estimates of themurder rate in the state than people who do not, but a majority ofFrederick’s respondents did not think of the city when questioned aboutthe state. Indeed, the average guess by people who were asked aboutMichigan is lower than the guesses of a similar group who were askedabout the murder rate in Detroit.

Blame for a failure to think of Detroit can be laid on both System 1 andSystem 2. Whether the city comes to mind when the state is mentioneddepends in part on the automatic function of memory. People differ in thisrespect. The representation of the state of Michigan is very detailed insome people’s minds: residents of the state are more likely to retrievemany facts about it than people who live elsewhere; geography buffs willretrieve more than others who specialize in baseball statistics; moreintelligent individuals are more likely than others to have richrepresentations of most things. Intelligence is not only the ability to reason;it is also the ability to find relevant material in memory and to deployattention when needed. Memory function is an attribute of System 1.However, everyone has the option of slowing down to conduct an activesearch of memory for all possibly relevant facts—just as they could slowdown to check the intuitive answer in the bat-and-ball problem. The extentof deliberate checking and search is a characteristic of System 2, whichvaries among individuals.

(PDF) Thinking, fast and slow by daniel kahneman - PDFSLIDE.US (48)

The bat-and-ball problem, the flowers syllogism, and theMichigan/Detroit problem have something in common. Failing theseminitests appears to be, at least to some extent, a matter of insufficientmotivation, not trying hard enough. Anyone who can be admitted to a gooduniversity is certainly able to reason through the first two questions and toreflect about Michigan long enough to remember the major city in that stateand its crime problem. These students can solve much more difficultproblems when they are not tempted to accept a superficially plausibleanswer that comes readily to mind. The ease with which they are satisfiedenough to stop thinking is rather troubling. “Lazy” is a harsh judgment aboutthe self-monitoring of these young people and their System 2, but it doesnot seem to be unfair. Those who avoid the sin of intellectual sloth could becalled “engaged.” They are more alert, more intellectually active, lesswilling to be satisfied with superficially attractive answers, more skepticalabout their intuitions. The psychologist Keith Stanovich would call themmore rational.

Intelligence, Control, Rationality

Researchers have applied diverse methods to examine the connectionbetween thinking and self-control. Some have addressed it by asking thecorrelation question: If people were ranked by their self-control and by theircognitive aptitude, would individuals have similar positions in the tworankings?

In one of the most famous experiments in the history of psychology,Walter Mischel and his students exposed four-year-old children to a crueldilemma. They were given a choice between a small reward (one Oreo),which they could have at any time, or a larger reward (two cookies) forwhich they had to wait 15 minutes under difficult conditions. They were toremain alone in a room, facing a desk with two objects: a single cookieand a bell that the child could ring at any time to call in the experimenterand receiven oand recei the one cookie. As the experiment wasdescribed: “There were no toys, books, pictures, or other potentiallydistracting items in the room. The experimenter left the room and did notreturn until 15 min had passed or the child had rung the bell, eaten therewards, stood up, or shown any signs of distress.”

The children were watched through a one-way mirror, and the film thatshows their behavior during the waiting time always has the audienceroaring in laughter. About half the children managed the feat of waiting for15 minutes, mainly by keeping their attention away from the temptingreward. Ten or fifteen years later, a large gap had opened between those

(PDF) Thinking, fast and slow by daniel kahneman - PDFSLIDE.US (49)

who had resisted temptation and those who had not. The resisters hadhigher measures of executive control in cognitive tasks, and especially theability to reallocate their attention effectively. As young adults, they wereless likely to take drugs. A significant difference in intellectual aptitudeemerged: the children who had shown more self-control as four-year-oldshad substantially higher scores on tests of intelligence.

A team of researchers at the University of Oregon explored the linkbetween cognitive control and intelligence in several ways, including anattempt to raise intelligence by improving the control of attention. Duringfive 40-minute sessions, they exposed children aged four to six to variouscomputer games especially designed to demand attention and control. Inone of the exercises, the children used a joystick to track a cartoon cat andmove it to a grassy area while avoiding a muddy area. The grassy areasgradually shrank and the muddy area expanded, requiring progressivelymore precise control. The testers found that training attention not onlyimproved executive control; scores on nonverbal tests of intelligence alsoimproved and the improvement was maintained for several months. Otherresearch by the same group identified specific genes that are involved inthe control of attention, showed that parenting techniques also affected thisability, and demonstrated a close connection between the children’s abilityto control their attention and their ability to control their emotions.

Shane Frederick constructed a Cognitive Reflection Test, whichconsists of the bat-and-ball problem and two other questions, chosenbecause they also invite an intuitive answer that is both compelling andwrong (the questions are shown here). He went on to study thecharacteristics of students who score very low on this test—the supervisoryfunction of System 2 is weak in these people—and found that they areprone to answer questions with the first idea that comes to mind andunwilling to invest the effort needed to check their intuitions. Individuals whouncritically follow their intuitions about puzzles are also prone to acceptother suggestions from System 1. In particular, they are impulsive,impatient, and keen to receive immediate gratification. For example, 63%of the intuitive respondents say they would prefer to get $3,400 this monthrather than $3,800 next month. Only 37% of those who solve all threepuzzles correctly have the same shortsighted preference for receiving asmaller amount immediately. When asked how much they will pay to getovernight delivery of a book they have ordered, the low scorers on theCognitive Reflection Test are willing to pay twice as much as the highscorers. Frederick’s findings suggest that the characters of ourpsychodrama have different “personalities.” System 1 is impulsive andintuitive; System 2 is capable of reasoning, and it is cautious, but at leastfor some people it is also lazy. We recognize related differences among

(PDF) Thinking, fast and slow by daniel kahneman - PDFSLIDE.US (50)

individuals: some people are more like their System 2; others are closer totheir System 1. This simple test has emerged as one of the betterpredictors of laztestors of ly thinking.

Keith Stanovich and his longtime collaborator Richard West originallyintroduced the terms System 1 and System 2 (they now prefer to speak ofType 1 and Type 2 processes). Stanovich and his colleagues have spentdecades studying differences among individuals in the kinds of problemswith which this book is concerned. They have asked one basic question inmany different ways: What makes some people more susceptible thanothers to biases of judgment? Stanovich published his conclusions in abook titled Rationality and the Reflective Mind, which offers a bold anddistinctive approach to the topic of this chapter. He draws a sharpdistinction between two parts of System 2—indeed, the distinction is sosharp that he calls them separate “minds.” One of these minds (he calls italgorithmic) deals with slow thinking and demanding computation. Somepeople are better than others in these tasks of brain power—they are theindividuals who excel in intelligence tests and are able to switch from onetask to another quickly and efficiently. However, Stanovich argues that highintelligence does not make people immune to biases. Another ability isinvolved, which he labels rationality. Stanovich’s concept of a rationalperson is similar to what I earlier labeled “engaged.” The core of hisargument is that rationality should be distinguished from intelligence. Inhis view, superficial or “lazy” thinking is a flaw in the reflective mind, afailure of rationality. This is an attractive and thought-provoking idea. Insupport of it, Stanovich and his colleagues have found that the bat-and-ballquestion and others like it are somewhat better indicators of oursusceptibility to cognitive errors than are conventional measures ofintelligence, such as IQ tests. Time will tell whether the distinction betweenintelligence and rationality can lead to new discoveries.

Speaking of Control

“She did not have to struggle to stay on task for hours. She was ina state of flow.”

“His ego was depleted after a long day of meetings. So he justturned to standard operating procedures instead of thinkingthrough the problem.”

(PDF) Thinking, fast and slow by daniel kahneman - PDFSLIDE.US (51)

“He didn’t bother to check whether what he said made sense.Does he usually have a lazy System 2 or was he unusually tired?”

“Unfortunately, she tends to say the first thing that comes into hermind. She probably also has trouble delaying gratification. WeakSystem 2.”

(PDF) Thinking, fast and slow by daniel kahneman - PDFSLIDE.US (52)

The Associative Machine

To begin your exploration of the surprising workings of System 1, look atthe following words:

Bananas Vomit

A lot happened to you during the last second or two. You experiencedsome unpleasant images and memories. Your face twisted slightly in anexpression of disgust, and you may have pushed this book imperceptiblyfarther away. Your heart rate increased, the hair on your arms rose a little,and your sweat glands were activated. In short, you responded to thedisgusting word with an attenuated version of how you would react to theactual event. All of this was completely automatic, beyond your control.

There was no particular reason to do so, but your mind automaticallyassumed a temporal sequence and a causal connection between thewords bananas and vomit, forming a sketchy scenario in which bananascaused the sickness. As a result, you are experiencing a temporaryaversion to bananas (don’t worry, it will pass). The state of your memoryhas changed in other ways: you are now unusually ready to recognize andrespond to objects and concepts associated with “vomit,” such as sick,stink, or nausea, and words associated with “bananas,” such as yellow andfruit, and perhaps apple and berries.

Vomiting normally occurs in specific contexts, such as hangovers andindigestion. You would also be unusually ready to recognize wordsassociated with other causes of the same unfortunate outcome.Furthermore, your System 1 noticed the fact that the juxtaposition of thetwo words is uncommon; you probably never encountered it before. Youexperienced mild surprise.

This complex constellation of responses occurred quickly, automatically,and effortlessly. You did not will it and you could not stop it. It was anoperation of System 1. The events that took place as a result of yourseeing the words happened by a process called associative activation:ideas that have been evoked trigger many other ideas, in a spreadingcascade of activity in your brain. The essential feature of this complex setof mental events is its coherence. Each element is connected, and eachsupports and strengthens the others. The word evokes memories, whichevoke emotions, which in turn evoke facial expressions and otherreactions, such as a general tensing up and an avoidance tendency. The

(PDF) Thinking, fast and slow by daniel kahneman - PDFSLIDE.US (53)

facial expression and the avoidance motion intensify the feelings to whichthey are linked, and the feelings in turn reinforce compatible ideas. All thishappens quickly and all at once, yielding a self-reinforcing pattern ofcognitive, emotional, and physical responses that is both diverse andintegrated—it has been called associatively coherent.

In a second or so you accomplished, automatically and unconsciously, aremarkable feat. Starting from a completely unexpected event, yourSystem 1 made as much sense as possible of the situation—two simplewords, oddly juxtaposed—by linking the words in a causal story; itevaluated the possible threat (mild to moderate) and created a context forfuture developments by preparing you for events that had just becomemore likely; it also created a context for the current event by evaluating howsurprising it was. You ended up as informed about the past and asprepared for the future as you could be.

An odd feature of what happened is that your System 1 treated the mereconjunction of two words as representations of reality. Your body reacted inan attenuated replica of a reaction to the real thing, and the emotionalresponse and physical recoil were part of the interpretation of the event. Ascognitive scientists have emphasized in recent years, cognition isembodied; you think with your body, not only with your brain.

The mechanism that causes these mental events has been known for along time: it is the ass12;velyociation of ideas. We all understand fromexperience that ideas follow each other in our conscious mind in a fairlyorderly way. The British philosophers of the seventeenth and eighteenthcenturies searched for the rules that explain such sequences. In AnEnquiry Concerning Human Understanding, published in 1748, theScottish philosopher David Hume reduced the principles of association tothree: resemblance, contiguity in time and place, and causality. Ourconcept of association has changed radically since Hume’s days, but histhree principles still provide a good start.

I will adopt an expansive view of what an idea is. It can be concrete orabstract, and it can be expressed in many ways: as a verb, as a noun, asan adjective, or as a clenched fist. Psychologists think of ideas as nodes ina vast network, called associative memory, in which each idea is linked tomany others. There are different types of links: causes are linked to theireffects (virus cold); things to their properties (lime green); things tothe categories to which they belong (banana fruit). One way we haveadvanced beyond Hume is that we no longer think of the mind as goingthrough a sequence of conscious ideas, one at a time. In the current viewof how associative memory works, a great deal happens at once. An ideathat has been activated does not merely evoke one other idea. It activates

(PDF) Thinking, fast and slow by daniel kahneman - PDFSLIDE.US (54)

many ideas, which in turn activate others. Furthermore, only a few of theactivated ideas will register in consciousness; most of the work ofassociative thinking is silent, hidden from our conscious selves. The notionthat we have limited access to the workings of our minds is difficult toaccept because, naturally, it is alien to our experience, but it is true: youknow far less about yourself than you feel you do.

The Marvels of Priming

As is common in science, the first big breakthrough in our understanding ofthe mechanism of association was an improvement in a method ofmeasurement. Until a few decades ago, the only way to study associationswas to ask many people questions such as, “What is the first word thatcomes to your mind when you hear the word DAY?” The researchers talliedthe frequency of responses, such as “night,” “sunny,” or “long.” In the 1980s,psychologists discovered that exposure to a word causes immediate andmeasurable changes in the ease with which many related words can beevoked. If you have recently seen or heard the word EAT, you aretemporarily more likely to complete the word fragment SO_P as SOUPthan as SOAP. The opposite would happen, of course, if you had just seenWASH. We call this a priming effect and say that the idea of EAT primesthe idea of SOUP, and that WASH primes SOAP.

Priming effects take many forms. If the idea of EAT is currently on yourmind (whether or not you are conscious of it), you will be quicker than usualto recognize the word SOUP when it is spoken in a whisper or presentedin a blurry font. And of course you are primed not only for the idea of soupbut also for a multitude of food-related ideas, including fork, hungry, fat,diet, and cookie. If for your most recent meal you sat at a wobbly restauranttable, you will be primed for wobbly as well. Furthermore, the primed ideashave some ability to prime other ideas, although more weakly. Like rippleson a pond, activation spreads through a small part of the vast network ofassociated ideas. The mapping of these ripples is now one of the mostexciting pursuits in psychological research.

Another major advance in our understanding of memory was thediscovery that priming is not restricted to concepts and words. You cannotknow this from conscious experience, of course, but you must accept thealien idea that your actions and your emotions can be primed by events ofwhich you are not even aware. In an experiment that became an instantclassic, the psychologist John Bargh and his collaborators asked studentsat New York University—most aged eighteen to twenty-two—to assemblefour-word sentences from a set of five words (for example, “finds he it

(PDF) Thinking, fast and slow by daniel kahneman - PDFSLIDE.US (55)

yellow instantly”). For one group of students, half the scrambled sentencescontained words associated with the elderly, such as Florida, forgetful,bald, gray, or wrinkle. When they had completed that task, the youngparticipants were sent out to do another experiment in an office down thehall. That short walk was what the experiment was about. The researchersunobtrusively measured the time it took people to get from one end of thecorridor to the other. As Bargh had predicted, the young people who hadfashioned a sentence from words with an elderly theme walked down thehallway significantly more slowly than the others.

The “Florida effect” involves two stages of priming. First, the set ofwords primes thoughts of old age, though the word old is never mentioned;second, these thoughts prime a behavior, walking slowly, which isassociated with old age. All this happens without any awareness. Whenthey were questioned afterward, none of the students reported noticing thatthe words had had a common theme, and they all insisted that nothing theydid after the first experiment could have been influenced by the words theyhad encountered. The idea of old age had not come to their consciousawareness, but their actions had changed nevertheless. This remarkablepriming phenomenon—the influencing of an action by the idea—is knownas the ideomotor effect. Although you surely were not aware of it, readingthis paragraph primed you as well. If you had needed to stand up to get aglass of water, you would have been slightly slower than usual to rise fromyour chair—unless you happen to dislike the elderly, in which caseresearch suggests that you might have been slightly faster than usual!

The ideomotor link also works in reverse. A study conducted in aGerman university was the mirror image of the early experiment that Barghand his colleagues had carried out in New York. Students were asked towalk around a room for 5 minutes at a rate of 30 steps per minute, whichwas about one-third their normal pace. After this brief experience, theparticipants were much quicker to recognize words related to old age,such as forgetful, old, and lonely. Reciprocal priming effects tend toproduce a coherent reaction: if you were primed to think of old age, youwould tend to act old, and acting old would reinforce the thought of old age.

Reciprocal links are common in the associative network. For example,being amused tends to make you smile, and smiling tends to make youfeel amused. Go ahead and take a pencil, and hold it between your teethfor a few seconds with the eraser pointing to your right and the point to yourleft. Now hold the pencil so the point is aimed straight in front of you, bypursing your lips around the eraser end. You were probably unaware thatone of these actions forced your face into a frown and the other into asmile. College students were asked to rate the humor of cartoons from

(PDF) Thinking, fast and slow by daniel kahneman - PDFSLIDE.US (56)

Gary Larson’s The Far Side while holding a pencil in their mouth. Thosewho were “smiling” (without any awareness of doing so) found the cartoonsrri221; (withfunnier than did those who were “frowning.” In anotherexperiment, people whose face was shaped into a frown (by squeezingtheir eyebrows together) reported an enhanced emotional response toupsetting pictures—starving children, people arguing, maimed accidentvictims.

Simple, common gestures can also unconsciously influence our thoughtsand feelings. In one demonstration, people were asked to listen tomessages through new headphones. They were told that the purpose ofthe experiment was to test the quality of the audio equipment and wereinstructed to move their heads repeatedly to check for any distortions ofsound. Half the participants were told to nod their head up and down whileothers were told to shake it side to side. The messages they heard wereradio editorials. Those who nodded (a yes gesture) tended to accept themessage they heard, but those who shook their head tended to reject it.Again, there was no awareness, just a habitual connection between anattitude of rejection or acceptance and its common physical expression.You can see why the common admonition to “act calm and kind regardlessof how you feel” is very good advice: you are likely to be rewarded byactually feeling calm and kind.

Primes That Guide Us

Studies of priming effects have yielded discoveries that threaten our self-image as conscious and autonomous authors of our judgments and ourchoices. For instance, most of us think of voting as a deliberate act thatreflects our values and our assessments of policies and is not influencedby irrelevancies. Our vote should not be affected by the location of thepolling station, for example, but it is. A study of voting patterns in precinctsof Arizona in 2000 showed that the support for propositions to increase thefunding of schools was significantly greater when the polling station was ina school than when it was in a nearby location. A separate experimentshowed that exposing people to images of classrooms and school lockersalso increased the tendency of participants to support a school initiative.The effect of the images was larger than the difference between parentsand other voters! The study of priming has come some way from the initialdemonstrations that reminding people of old age makes them walk moreslowly. We now know that the effects of priming can reach into every cornerof our lives.

Reminders of money produce some troubling effects. Participants in one

(PDF) Thinking, fast and slow by daniel kahneman - PDFSLIDE.US (57)

experiment were shown a list of five words from which they were requiredto construct a four-word phrase that had a money theme (“high a salarydesk paying” became “a high-paying salary”). Other primes were muchmore subtle, including the presence of an irrelevant money-related objectin the background, such as a stack of Monopoly money on a table, or acomputer with a screen saver of dollar bills floating in water.

Money-primed people become more independent than they would bewithout the associative trigger. They persevered almost twice as long intrying to solve a very difficult problem before they asked the experimenterfor help, a crisp demonstration of increased self-reliance. Money-primedpeople are also more selfish: they were much less willing to spend timehelping another student who pretended to be confused about anexperimental task. When an experimenter clumsily dropped a bunch ofpencils on the floor, the participants with money (unconsciously) on theirmind picked up fewer pencils. In another experiment in the series,participants were told that they would shortly have a get-acquaintedconversation with another person and were asked to set up two chairswhile the experimenter left to retrieve that person. Participants primed bymoney chose in the exto stay much farther apart than their nonprimedpeers (118 vs. 80 centimeters). Money-primed undergraduates alsoshowed a greater preference for being alone.

The general theme of these findings is that the idea of money primesindividualism: a reluctance to be involved with others, to depend on others,or to accept demands from others. The psychologist who has done thisremarkable research, Kathleen Vohs, has been laudably restrained indiscussing the implications of her findings, leaving the task to her readers.Her experiments are profound—her findings suggest that living in a culturethat surrounds us with reminders of money may shape our behavior andour attitudes in ways that we do not know about and of which we may notbe proud. Some cultures provide frequent reminders of respect, othersconstantly remind their members of God, and some societies primeobedience by large images of the Dear Leader. Can there be any doubtthat the ubiquitous portraits of the national leader in dictatorial societiesnot only convey the feeling that “Big Brother Is Watching” but also lead toan actual reduction in spontaneous thought and independent action?

The evidence of priming studies suggests that reminding people of theirmortality increases the appeal of authoritarian ideas, which may becomereassuring in the context of the terror of death. Other experiments haveconfirmed Freudian insights about the role of symbols and metaphors inunconscious associations. For example, consider the ambiguous wordfragments W_ _ H and S_ _ P. People who were recently asked to think ofan action of which they are ashamed are more likely to complete those

(PDF) Thinking, fast and slow by daniel kahneman - PDFSLIDE.US (58)

fragments as WASH and SOAP and less likely to see WISH and SOUP.Furthermore, merely thinking about stabbing a coworker in the back leavespeople more inclined to buy soap, disinfectant, or detergent than batteries,juice, or candy bars. Feeling that one’s soul is stained appears to trigger adesire to cleanse one’s body, an impulse that has been dubbed the “LadyMacbeth effect.”

The cleansing is highly specific to the body parts involved in a sin.Participants in an experiment were induced to “lie” to an imaginary person,either on the phone or in e-mail. In a subsequent test of the desirability ofvarious products, people who had lied on the phone preferred mouthwashover soap, and those who had lied in e-mail preferred soap to mouthwash.

When I describe priming studies to audiences, the reaction is oftendisbelief. This is not a surprise: System 2 believes that it is in charge andthat it knows the reasons for its choices. Questions are probably croppingup in your mind as well: How is it possible for such trivial manipulations ofthe context to have such large effects? Do these experiments demonstratethat we are completely at the mercy of whatever primes the environmentprovides at any moment? Of course not. The effects of the primes arerobust but not necessarily large. Among a hundred voters, only a fewwhose initial preferences were uncertain will vote differently about a schoolissue if their precinct is located in a school rather than in a church—but afew percent could tip an election.

The idea you should focus on, however, is that disbelief is not an option.The results are not made up, nor are they statistical flukes. You have nochoice but to accept that the major conclusions of these studies are true.More important, you must accept that they are true about you. If you hadbeen exposed to a screen saver of floating dollar bills, you too would likelyhave picked up fewer pencils to help a clumsy stranger. You do not believethat these results apply to you because they correspond to nothing in yoursubjective experience. But your subjective expefteelief. Trience consistslargely of the story that your System 2 tells itself about what is going on.Priming phenomena arise in System 1, and you have no conscious accessto them.

I conclude with a perfect demonstration of a priming effect, which wasconducted in an office kitchen at a British university. For many yearsmembers of that office had paid for the tea or coffee to which they helpedthemselves during the day by dropping money into an “honesty box.” A listof suggested prices was posted. One day a banner poster was displayedjust above the price list, with no warning or explanation. For a period of tenweeks a new image was presented each week, either flowers or eyes thatappeared to be looking directly at the observer. No one commented on the

(PDF) Thinking, fast and slow by daniel kahneman - PDFSLIDE.US (59)

new decorations, but the contributions to the honesty box changedsignificantly. The posters and the amounts that people put into the cashbox (relative to the amount they consumed) are shown in figure 4. Theydeserve a close look.

Figure 4

On the first week of the experiment (which you can see at the bottom of thefigure), two wide-open eyes stare at the coffee or tea drinkers, whoseaverage contribution was 70 pence per liter of milk. On week 2, the postershows flowers and average contributions drop to about 15 pence. Thetrend continues. On average, the users of the kitchen contributed almostthree times as much in “eye weeks” as they did in “flower weeks.”Evidently, a purely symbolic reminder of being watched prodded peopleinto improved behavior. As we expect at this point, the effect occurswithout any awareness. Do you now believe that you would also fall into thesame pattern?

Some years ago, the psychologist Timothy Wilson wrote a book with theevocative title Strangers to Ourselves. You have now been introduced tothat stranger in you, which may be in control of much of what you do,although you rarely have a glimpse of it. System 1 provides theimpressions that often turn into your beliefs, and is the source of theimpulses that often become your choices and your actions. It offers a tacitinterpretation of what happens to you and around you, linking the present

(PDF) Thinking, fast and slow by daniel kahneman - PDFSLIDE.US (60)

with the recent past and with expectations about the near future. It containsthe model of the world that instantly evaluates events as normal orsurprising. It is the source of your rapid and often precise intuitivejudgments. And it does most of this without your conscious awareness ofits activities. System 1 is also, as we will see in the following chapters, theorigin of many of the systematic errors in your intuitions.

Speaking of Priming

“The sight of all these people in uniforms does not primecreativity.”

“The world makes much less sense than you think. Thecoherence comes mostly from the way your mind works.”

“They were primed to find flaws, and this is exactly what theyfound.”

“His System 1 constructed a story, and his System 2 believed it. Ithappens to allel

“I made myself smile and I’m actually feeling better!”

(PDF) Thinking, fast and slow by daniel kahneman - PDFSLIDE.US (61)

Cognitive Ease

Whenever you are conscious, and perhaps even when you are not, multiplecomputations are going on in your brain, which maintain and updatecurrent answers to some key questions: Is anything new going on? Is therea threat? Are things going well? Should my attention be redirected? Ismore effort needed for this task? You can think of a cockpit, with a set ofdials that indicate the current values of each of these essential variables.The assessments are carried out automatically by System 1, and one oftheir functions is to determine whether extra effort is required from System2.

One of the dials measures cognitive ease, and its range is between“Easy” and “Strained.” Easy is a sign that things are going well—nothreats, no major news, no need to redirect attention or mobilize effort.Strained indicates that a problem exists, which will require increasedmobilization of System 2. Conversely, you experience cognitive strain.Cognitive strain is affected by both the current level of effort and thepresence of unmet demands. The surprise is that a single dial of cognitiveease is connected to a large network of diverse inputs and outputs. Figure5 tells the story.

The figure suggests that a sentence that is printed in a clear font, or hasbeen repeated, or has been primed, will be fluently processed withcognitive ease. Hearing a speaker when you are in a good mood, or evenwhen you have a pencil stuck crosswise in your mouth to make you “smile,”also induces cognitive ease. Conversely, you experience cognitive strainwhen you read instructions in a poor font, or in faint colors, or worded incomplicated language, or when you are in a bad mood, and even when youfrown.

Figure 5. Causes and Consequences ofCognitive Ease

(PDF) Thinking, fast and slow by daniel kahneman - PDFSLIDE.US (62)

The various causes of ease or strain have interchangeable effects.When you are in a state of cognitive ease, you are probably in a goodmood, like what you see, believe what you hear, trust your intuitions, andfeel that the current situation is comfortably familiar. You are also likely tobe relatively casual and superficial in your thinking. When you feel strained,you are more likely to be vigilant and suspicious, invest more effort in whatyou are doing, feel less comfortable, and make fewer errors, but you alsoare less intuitive and less creative than usual.

Illusions of Remembering

The word illusion brings visual illusions to mind, because we are allfamiliar with pictures that mislead. But vision is not the only domain ofillusions; memory is also susceptible to them, as is thinking moregenerally.

David Stenbill, Monica Bigoutski, Sh"imight=s is pictana Tirana. I justmade up these names. If you encounter any of them within the next fewminutes you are likely to remember where you saw them. You know, andwill know for a while, that these are not the names of minor celebrities. Butsuppose that a few days from now you are shown a long list of names,including some minor celebrities and “new” names of people that you havenever heard of; your task will be to check every name of a celebrity in thelist. There is a substantial probability that you will identify David Stenbill asa well-known person, although you will not (of course) know whether youencountered his name in the context of movies, sports, or politics. LarryJacoby, the psychologist who first demonstrated this memory illusion in thelaboratory, titled his article “Becoming Famous Overnight.” How does thishappen? Start by asking yourself how you know whether or not someone isfamous. In some cases of truly famous people (or of celebrities in an areayou follow), you have a mental file with rich information about a person—think Albert Einstein, Bono, Hillary Clinton. But you will have no file ofinformation about David Stenbill if you encounter his name in a few days.All you will have is a sense of familiarity—you have seen this namesomewhere.

Jacoby nicely stated the problem: “The experience of familiarity has asimple but powerful quality of ‘pastness’ that seems to indicate that it is adirect reflection of prior experience.” This quality of pastness is an illusion.The truth is, as Jacoby and many followers have shown, that the nameDavid Stenbill will look familiar when you see it because you will see itmore clearly. Words that you have seen before become easier to see

(PDF) Thinking, fast and slow by daniel kahneman - PDFSLIDE.US (63)

again—you can identify them better than other words when they are shownvery briefly or masked by noise, and you will be quicker (by a fewhundredths of a second) to read them than to read other words. In short,you experience greater cognitive ease in perceiving a word you have seenearlier, and it is this sense of ease that gives you the impression offamiliarity.

Figure 5 suggests a way to test this. Choose a completely new word,make it easier to see, and it will be more likely to have the quality ofpastness. Indeed, a new word is more likely to be recognized as familiar ifit is unconsciously primed by showing it for a few milliseconds just beforethe test, or if it is shown in sharper contrast than some other words in thelist. The link also operates in the other direction. Imagine you are shown alist of words that are more or less out of focus. Some of the words areseverely blurred, others less so, and your task is to identify the words thatare shown more clearly. A word that you have seen recently will appear tobe clearer than unfamiliar words. As figure 5 indicates, the various ways ofinducing cognitive ease or strain are interchangeable; you may not knowprecisely what it is that makes things cognitively easy or strained. This ishow the illusion of familiarity comes about.

Illusions of Truth

“New York is a large city in the United States.” “The moon revolves aroundEarth.” “A chicken has four legs.” In all these cases, you quickly retrieved agreat deal of related information, almost all pointing one way or another.You knew soon after reading them that the first two statements are true andthe last one is false. Note, however, that the statement “A chicken hasthree legs” is more obviously false than “A chicken has four legs.” Yourassociative machinery slows the judgment of the latter sentence bydelivering the fact that many animals have four legs, and perhaps also thatsupermarkets often sell chickenordblurred, legs in packages of four.System 2 was involved in sifting that information, perhaps raising the issueof whether the question about New York was too easy, or checking themeaning of revolves.

Think of the last time you took a driving test. Is it true that you need aspecial license to drive a vehicle that weighs more than three tons?Perhaps you studied seriously and can remember the side of the page onwhich the answer appeared, as well as the logic behind it. This is certainlynot how I passed driving tests when I moved to a new state. My practicewas to read the booklet of rules quickly once and hope for the best. I knewsome of the answers from the experience of driving for a long time. But

(PDF) Thinking, fast and slow by daniel kahneman - PDFSLIDE.US (64)

there were questions where no good answer came to mind, where all I hadto go by was cognitive ease. If the answer felt familiar, I assumed that itwas probably true. If it looked new (or improbably extreme), I rejected it.The impression of familiarity is produced by System 1, and System 2relies on that impression for a true/false judgment.

The lesson of figure 5 is that predictable illusions inevitably occur if ajudgment is based on an impression of cognitive ease or strain. Anythingthat makes it easier for the associative machine to run smoothly will alsobias beliefs. A reliable way to make people believe in falsehoods isfrequent repetition, because familiarity is not easily distinguished fromtruth. Authoritarian institutions and marketers have always known this fact.But it was psychologists who discovered that you do not have to repeat theentire statement of a fact or idea to make it appear true. People who wererepeatedly exposed to the phrase “the body temperature of a chicken”were more likely to accept as true the statement that “the body temperatureof a chicken is 144°” (or any other arbitrary number). The familiarity of onephrase in the statement sufficed to make the whole statement feel familiar,and therefore true. If you cannot remember the source of a statement, andhave no way to relate it to other things you know, you have no option but togo with the sense of cognitive ease.

How to Write a Persuasive Message

Suppose you must write a message that you want the recipients to believe.Of course, your message will be true, but that is not necessarily enough forpeople to believe that it is true. It is entirely legitimate for you to enlistcognitive ease to work in your favor, and studies of truth illusions providespecific suggestions that may help you achieve this goal.

The general principle is that anything you can do to reduce cognitivestrain will help, so you should first maximize legibility. Compare these twostatements:

Adolf Hitler was born in 1892.Adolf Hitler was born in 1887.

Both are false (Hitler was born in 1889), but experiments have shown thatthe first is more likely to be believed. More advice: if your message is to beprinted, use high-quality paper to maximize the contrast betweencharacters and their background. If you use color, you are more likely to bebelieved if your text is printed in bright blue or red than in middling shadesof green, yellow, or pale blue.

(PDF) Thinking, fast and slow by daniel kahneman - PDFSLIDE.US (65)

If you care about being thought credible and intelligent, do not usecomplex language where simpler language will do. My Princeton toncolleague Danny Oppenheimer refuted a myth prevalent a wo ton colmongundergraduates about the vocabulary that professors find most impressive.In an article titled “Consequences of Erudite Vernacular UtilizedIrrespective of Necessity: Problems with Using Long Words Needlessly,”he showed that couching familiar ideas in pretentious language is taken asa sign of poor intelligence and low credibility.

In addition to making your message simple, try to make it memorable.Put your ideas in verse if you can; they will be more likely to be taken astruth. Participants in a much cited experiment read dozens of unfamiliaraphorisms, such as:

Woes unite foes.Little strokes will tumble great oaks.A fault confessed is half redressed.

Other students read some of the same proverbs transformed intononrhyming versions:

Woes unite enemies.Little strokes will tumble great trees.A fault admitted is half redressed.

The aphorisms were judged more insightful when they rhymed than whenthey did not.

Finally, if you quote a source, choose one with a name that is easy topronounce. Participants in an experiment were asked to evaluate theprospects of fictitious Turkish companies on the basis of reports from twobrokerage firms. For each stock, one of the reports came from an easilypronounced name (e.g., Artan) and the other report came from a firm withan unfortunate name (e.g., Taahhut). The reports sometimes disagreed.The best procedure for the observers would have been to average the tworeports, but this is not what they did. They gave much more weight to thereport from Artan than to the report from Taahhut. Remember that System2 is lazy and that mental effort is aversive. If possible, the recipients of yourmessage want to stay away from anything that reminds them of effort,including a source with a complicated name.

All this is very good advice, but we should not get carried away. High-quality paper, bright colors, and rhyming or simple language will not bemuch help if your message is obviously nonsensical, or if it contradictsfacts that your audience knows to be true. The psychologists who do these

(PDF) Thinking, fast and slow by daniel kahneman - PDFSLIDE.US (66)

experiments do not believe that people are stupid or infinitely gullible. Whatpsychologists do believe is that all of us live much of our life guided by theimpressions of System 1—and we often do not know the source of theseimpressions. How do you know that a statement is true? If it is stronglylinked by logic or association to other beliefs or preferences you hold, orcomes from a source you trust and like, you will feel a sense of cognitiveease. The trouble is that there may be other causes for your feeling of ease—including the quality of the font and the appealing rhythm of the prose—and you have no simple way of tracing your feelings to their source. This isthe message of figure 5: the sense of ease or strain has multiple causes,and it is difficult to tease them apart. Difficult, but not impossible. Peoplecan overcome some of the superficial factors that produce illusions of truthwhen strongly motivated to do so. On most occasions, however, the lazySystem 2 will adopt the suggestions of System 1 and march on.

Strain and Effort

The symmetry of many associative connections was a dominant theme inthe discussion of associative coherence. As we saw earlier, people whoare made to “smile” or “frown” by sticking a pencil in their mouth or holdinga ball between their furrowed brows are prone to experience the emotionsthat frowning and smiling normally express. The same self-reinforcingreciprocity is found in studies of cognitive ease. On the one hand, cognitivestrain is experienced when the effortful operations of System 2 areengaged. On the other hand, the experience of cognitive strain, whateverits source, tends to mobilize System 2, shifting people’s approach toproblems from a casual intuitive mode to a more engaged and analyticmode.

The bat-and-ball problem was mentioned earlier as a test of people’stendency to answer questions with the first idea that comes to their mind,without checking it. Shane Frederick’s Cognitive Reflection Test consistsof the bat-and-ball problem and two others, all chosen because they evokean immediate intuitive answer that is incorrect. The other two items in theCRT are:

If it takes 5 machines 5 minutes to make 5 widgets, how longwould it take 100 machines to make 100 widgets?

100 minutes OR 5 minutes

In a lake, there is a patch of lily pads. Every day, the patchdoubles in size.

(PDF) Thinking, fast and slow by daniel kahneman - PDFSLIDE.US (67)

If it takes 48 days for the patch to cover the entire lake, how longwould it take for the patch to cover half of the lake?

24 days OR 47 days

The correct answers to both problems are in a footnote at the bottom of thepage.* The experimenters recruited 40 Princeton students to take the CRT.Half of them saw the puzzles in a small font in washed-out gray print. Thepuzzles were legible, but the font induced cognitive strain. The results tell aclear story: 90% of the students who saw the CRT in normal font made atleast one mistake in the test, but the proportion dropped to 35% when thefont was barely legible. You read this correctly: performance was betterwith the bad font. Cognitive strain, whatever its source, mobilizes System2, which is more likely to reject the intuitive answer suggested by System1.

The Pleasure of Cognitive Ease

An article titled “Mind at Ease Puts a Smile on the Face” describes anexperiment in which participants were briefly shown pictures of objects.Some of these pictures were made easier to recognize by showing theoutline of the object just before the complete image was shown, so brieflythat the contours were never noticed. Emotional reactions were measuredby recording electrical impulses from facial muscles, registering changesof expression that are too slight and too brief to be detectable byobservers. As expected, people showed a faint smile and relaxed browswhen the pictures were easier to see. It appears to be a feature of System1 that cognitive ease is associated with good feelings.

As expected, easily pronounced words evoke a favorable attitude.Companies with pronounceable names dmisorrectlo better than others forthe first week after the stock is issued, though the effect disappears overtime. Stocks with pronounceable trading symbols (like KAR or LUNMOO)outperform those with tongue-twisting tickers like PXG or RDO—and theyappear to retain a small advantage over some time. A study conducted inSwitzerland found that investors believe that stocks with fluent names likeEmmi, Swissfirst, and Comet will earn higher returns than those with clunkylabels like Geberit and Ypsomed.

As we saw in figure 5, repetition induces cognitive ease and acomforting feeling of familiarity. The famed psychologist Robert Zajoncdedicated much of his career to the study of the link between the repetitionof an arbitrary stimulus and the mild affection that people eventually havefor it. Zajonc called it the mere exposure effect. A demonstration

(PDF) Thinking, fast and slow by daniel kahneman - PDFSLIDE.US (68)

conducted in the student newspapers of the University of Michigan and ofMichigan State University is one of my favorite experiments. For a periodof some weeks, an ad-like box appeared on the front page of the paper,which contained one of the following Turkish (or Turkish-sounding) words:kadirga, saricik, biwonjni, nansoma, and iktitaf. The frequency with whichthe words were repeated varied: one of the words was shown only once,the others appeared on two, five, ten, or twenty-five separate occasions.(The words that were presented most often in one of the university paperswere the least frequent in the other.) No explanation was offered, andreaders’ queries were answered by the statement that “the purchaser ofthe display wished for anonymity.”

When the mysterious series of ads ended, the investigators sentquestionnaires to the university communities, asking for impressions ofwhether each of the words “means something ‘good’ or something ‘bad.’”The results were spectacular: the words that were presented morefrequently were rated much more favorably than the words that had beenshown only once or twice. The finding has been confirmed in manyexperiments, using Chinese ideographs, faces, and randomly shapedpolygons.

The mere exposure effect does not depend on the consciousexperience of familiarity. In fact, the effect does not depend onconsciousness at all: it occurs even when the repeated words or picturesare shown so quickly that the observers never become aware of havingseen them. They still end up liking the words or pictures that werepresented more frequently. As should be clear by now, System 1 canrespond to impressions of events of which System 2 is unaware. Indeed,the mere exposure effect is actually stronger for stimuli that the individualnever consciously sees.

Zajonc argued that the effect of repetition on liking is a profoundlyimportant biological fact, and that it extends to all animals. To survive in afrequently dangerous world, an organism should react cautiously to a novelstimulus, with withdrawal and fear. Survival prospects are poor for ananimal that is not suspicious of novelty. However, it is also adaptive for theinitial caution to fade if the stimulus is actually safe. The mere exposureeffect occurs, Zajonc claimed, because the repeated exposure of astimulus is followed by nothing bad. Such a stimulus will eventually becomea safety signal, and safety is good. Obviously, this argument is notrestricted to humans. To make that point, one of Zajonc’s associatesexposed two sets of fertile chicken eggs to different tones. After theyhatched, the chicks consistently emitted fewer distress calls when exposedto the tone they had heard while inhabiting the shell.

Zajonc offered an eloquent summary of hing icts program of research:

(PDF) Thinking, fast and slow by daniel kahneman - PDFSLIDE.US (69)

Zajonc offered an eloquent summary of hing icts program of research:

The consequences of repeated exposures benefit the organismin its relations to the immediate animate and inanimateenvironment. They allow the organism to distinguish objects andhabitats that are safe from those that are not, and they are themost primitive basis of social attachments. Therefore, they formthe basis for social organization and cohesion—the basicsources of psychological and social stability.

The link between positive emotion and cognitive ease in System 1 has along evolutionary history.

Ease, Mood, and Intuition

Around 1960, a young psychologist named Sarnoff Mednick thought hehad identified the essence of creativity. His idea was as simple as it waspowerful: creativity is associative memory that works exceptionally well. Hemade up a test, called the Remote Association Test (RAT), which is stilloften used in studies of creativity.

For an easy example, consider the following three words:cottage Swiss cake

Can you think of a word that is associated with all three? You probablyworked out that the answer is cheese. Now try this:

dive light rocketThis problem is much harder, but it has a unique correct answer, whichevery speaker of English recognizes, although less than 20% of a sampleof students found it within 15 seconds. The answer is sky. Of course, notevery triad of words has a solution. For example, the words dream, ball,book do not have a shared association that everyone will recognize asvalid.

Several teams of German psychologists that have studied the RAT inrecent years have come up with remarkable discoveries about cognitiveease. One of the teams raised two questions: Can people feel that a triadof words has a solution before they know what the solution is? How doesmood influence performance in this task? To find out, they first made someof their subjects happy and others sad, by asking them to think for severalminutes about happy or sad episodes in their lives. Then they presentedthese subjects with a series of triads, half of them linked (such as dive,light, rocket) and half unlinked (such as dream, ball, book), and instructedthem to press one of two keys very quickly to indicate their guess aboutwhether the triad was linked. The time allowed for this guess, 2 seconds,

(PDF) Thinking, fast and slow by daniel kahneman - PDFSLIDE.US (70)

was much too short for the actual solution to come to anyone’s mind.The first surprise is that people’s guesses are much more accurate than

they would be by chance. I find this astonishing. A sense of cognitive easeis apparently generated by a very faint signal from the associativemachine, which “knows” that the three words are coherent (share anassociation) long before the association is retrieved. The role of cognitiveease in the judgment was confirmed experimentally by another Germanteam: manipulations that increase cognitive ease (priming, a clear font,pre-exposing words) all increase the tendency to see the words as linked.

Another remarkable discovery is the powerful effect of mood on thisintuitive performance. The experimentershape tende computed an“intuition index” to measure accuracy. They found that putting theparticipants in a good mood before the test by having them think happythoughts more than doubled accuracy. An even more striking result is thatunhappy subjects were completely incapable of performing the intuitivetask accurately; their guesses were no better than random. Mood evidentlyaffects the operation of System 1: when we are uncomfortable andunhappy, we lose touch with our intuition.

These findings add to the growing evidence that good mood, intuition,creativity, gullibility, and increased reliance on System 1 form a cluster. Atthe other pole, sadness, vigilance, suspicion, an analytic approach, andincreased effort also go together. A happy mood loosens the control ofSystem 2 over performance: when in a good mood, people become moreintuitive and more creative but also less vigilant and more prone to logicalerrors. Here again, as in the mere exposure effect, the connection makesbiological sense. A good mood is a signal that things are generally goingwell, the environment is safe, and it is all right to let one’s guard down. Abad mood indicates that things are not going very well, there may be athreat, and vigilance is required. Cognitive ease is both a cause and aconsequence of a pleasant feeling.

The Remote Association Test has more to tell us about the link betweencognitive ease and positive affect. Briefly consider two triads of words:

sleep mail switchsalt deep foam

You could not know it, of course, but measurements of electrical activity inthe muscles of your face would probably have shown a slight smile whenyou read the second triad, which is coherent (sea is the solution). Thissmiling reaction to coherence appears in subjects who are told nothingabout common associates; they are merely shown a vertically arrangedtriad of words and instructed to press the space bar after they have read it.The impression of cognitive ease that comes with the presentation of acoherent triad appears to be mildly pleasurable in itself.

(PDF) Thinking, fast and slow by daniel kahneman - PDFSLIDE.US (71)

coherent triad appears to be mildly pleasurable in itself.The evidence that we have about good feelings, cognitive ease, and the

intuition of coherence is, as scientists say, correlational but not necessarilycausal. Cognitive ease and smiling occur together, but do the goodfeelings actually lead to intuitions of coherence? Yes, they do. The proofcomes from a clever experimental approach that has become increasinglypopular. Some participants were given a cover story that provided analternative interpretation for their good feeling: they were told about musicplayed in their earphones that “previous research showed that this musicinfluences the emotional reactions of individuals.” This story completelyeliminates the intuition of coherence. The finding shows that the briefemotional response that follows the presentation of a triad of words(pleasant if the triad is coherent, unpleasant otherwise) is actually the basisof judgments of coherence. There is nothing here that System 1 cannot do.Emotional changes are now expected, and because they are unsurprisingthey are not linked causally to the words.

This is as good as psychological research ever gets, in its combinationof experimental techniques and in its results, which are both robust andextremely surprising. We have learned a great deal about the automaticworkings of System 1 in the last decades. Much of what we now knowwould have sounded like science fiction thirty or forty years ago. It wasbeyond imagining that bad font influences judgments of truth and improvescognitive performance, or that an emotional response to the cognitiveease of a tri pr that aad of words mediates impressions of coherence.Psychology has come a long way.

Speaking of Cognitive Ease

“Let’s not dismiss their business plan just because the fontmakes it hard to read.”

“We must be inclined to believe it because it has been repeatedso often, but let’s think it through again.”

“Familiarity breeds liking. This is a mere exposure effect.”

“I’m in a very good mood today, and my System 2 is weaker thanusual. I should be extra careful.”

(PDF) Thinking, fast and slow by daniel kahneman - PDFSLIDE.US (72)

Norms, Surprises, and Causes

The central characteristics and functions of System 1 and System 2 havenow been introduced, with a more detailed treatment of System 1. Freelymixing metaphors, we have in our head a remarkably powerful computer,not fast by conventional hardware standards, but able to represent thestructure of our world by various types of associative links in a vast networkof various types of ideas. The spreading of activation in the associativemachine is automatic, but we (System 2) have some ability to control thesearch of memory, and also to program it so that the detection of an eventin the environment can attract attention. We next go into more detail of thewonders and limitation of what System 1 can do.

Assessing Normality

The main function of System 1 is to maintain and update a model of yourpersonal world, which represents what is normal in it. The model isconstructed by associations that link ideas of circumstances, events,actions, and outcomes that co-occur with some regularity, either at thesame time or within a relatively short interval. As these links are formedand strengthened, the pattern of associated ideas comes to represent thestructure of events in your life, and it determines your interpretation of thepresent as well as your expectations of the future.

A capacity for surprise is an essential aspect of our mental life, andsurprise itself is the most sensitive indication of how we understand ourworld and what we expect from it. There are two main varieties of surprise.Some expectations are active and conscious—you know you are waitingfor a particular event to happen. When the hour is near, you may beexpecting the sound of the door as your child returns from school; when thedoor opens you expect the sound of a familiar voice. You will be surprisedif an actively expected event does not occur. But there is a much largercategory of events that you expect passively; you don’t wait for them, butyou are not surprised when they happen. These are events that are normalin a situation, though not sufficiently probable to be actively expected.

A single incident may make a recurrence less surprising. Some yearsago, my wife and I were of dealWhen normvacationing in a small islandresort on the Great Barrier Reef. There are only forty guest rooms on theisland. When we came to dinner, we were surprised to meet anacquaintance, a psychologist named Jon. We greeted each other warmlyand commented on the coincidence. Jon left the resort the next day. Abouttwo weeks later, we were in a theater in London. A latecomer sat next to

(PDF) Thinking, fast and slow by daniel kahneman - PDFSLIDE.US (73)

me after the lights went down. When the lights came up for theintermission, I saw that my neighbor was Jon. My wife and I commentedlater that we were simultaneously conscious of two facts: first, this was amore remarkable coincidence than the first meeting; second, we weredistinctly less surprised to meet Jon on the second occasion than we hadbeen on the first. Evidently, the first meeting had somehow changed theidea of Jon in our minds. He was now “the psychologist who shows upwhen we travel abroad.” We (System 2) knew this was a ludicrous idea,but our System 1 had made it seem almost normal to meet Jon in strangeplaces. We would have experienced much more surprise if we had metany acquaintance other than Jon in the next seat of a London theater. Byany measure of probability, meeting Jon in the theater was much less likelythan meeting any one of our hundreds of acquaintances—yet meeting Jonseemed more normal.

Under some conditions, passive expectations quickly turn active, as wefound in another coincidence. On a Sunday evening some years ago, wewere driving from New York City to Princeton, as we had been doing everyweek for a long time. We saw an unusual sight: a car on fire by the side ofthe road. When we reached the same stretch of road the following Sunday,another car was burning there. Here again, we found that we were distinctlyless surprised on the second occasion than we had been on the first. Thiswas now “the place where cars catch fire.” Because the circumstances ofthe recurrence were the same, the second incident was sufficient to createan active expectation: for months, perhaps for years, after the event wewere reminded of burning cars whenever we reached that spot of the roadand were quite prepared to see another one (but of course we never did).

The psychologist Dale Miller and I wrote an essay in which we attemptedto explain how events come to be perceived as normal or abnormal. I willuse an example from our description of “norm theory,” although myinterpretation of it has changed slightly:

An observer, casually watching the patrons at a neighboring tablein a fashionable restaurant, notices that the first guest to taste thesoup winces, as if in pain. The normality of a multitude of eventswill be altered by this incident. It is now unsurprising for the guestwho first tasted the soup to startle violently when touched by awaiter; it is also unsurprising for another guest to stifle a cry whentasting soup from the same tureen. These events and manyothers appear more normal than they would have otherwise, butnot necessarily because they confirm advance expectations.Rather, they appear normal because they recruit the originalepisode, retrieve it from memory, and are interpreted in

(PDF) Thinking, fast and slow by daniel kahneman - PDFSLIDE.US (74)

conjunction with it.

Imagine yourself the observer at the restaurant. You were surprised bythe first guest’s unusual reaction to the soup, and surprised again by thestartled response to the waiter’s touch. However, the second abnormalevent will retrieve the first from memory, and both make sense together.The two events fit into a pattern, in which the guest is an exceptionallytense person. On the other hand, if the next thing that happens after the firstguest’s grimace is that another customer rejects the soup, these twosurprises will be linked and thehinsur soup will surely be blamed.

“How many animals of each kind did Moses take into the ark?” Thenumber of people who detect what is wrong with this question is so smallthat it has been dubbed the “Moses illusion.” Moses took no animals intothe ark; Noah did. Like the incident of the wincing soup eater, the Mosesillusion is readily explained by norm theory. The idea of animals going intothe ark sets up a biblical context, and Moses is not abnormal in thatcontext. You did not positively expect him, but the mention of his name isnot surprising. It also helps that Moses and Noah have the same vowelsound and number of syllables. As with the triads that produce cognitiveease, you unconsciously detect associative coherence between “Moses”and “ark” and so quickly accept the question. Replace Moses with GeorgeW. Bush in this sentence and you will have a poor political joke but noillusion.

When something cement does not fit into the current context of activatedideas, the system detects an abnormality, as you just experienced. Youhad no particular idea of what was coming after something, but you knewwhen the word cement came that it was abnormal in that sentence.Studies of brain responses have shown that violations of normality aredetected with astonishing speed and subtlety. In a recent experiment,people heard the sentence “Earth revolves around the trouble every year.”A distinctive pattern was detected in brain activity, starting within two-tenths of a second of the onset of the odd word. Even more remarkable,the same brain response occurs at the same speed when a male voicesays, “I believe I am pregnant because I feel sick every morning,” or whenan upper-class voice says, “I have a large tattoo on my back.” A vastamount of world knowledge must instantly be brought to bear for theincongruity to be recognized: the voice must be identified as upper-classEnglish and confronted with the generalization that large tattoos areuncommon in the upper class.

We are able to communicate with each other because our knowledge ofthe world and our use of words are largely shared. When I mention a table,

(PDF) Thinking, fast and slow by daniel kahneman - PDFSLIDE.US (75)

without specifying further, you understand that I mean a normal table. Youknow with certainty that its surface is approximately level and that it has farfewer than 25 legs. We have norms for a vast number of categories, andthese norms provide the background for the immediate detection ofanomalies such as pregnant men and tattooed aristocrats.

To appreciate the role of norms in communication, consider thesentence “The large mouse climbed over the trunk of the very smallelephant.” I can count on your having norms for the size of mice andelephants that are not too far from mine. The norms specify a typical oraverage size for these animals, and they also contain information about therange or variability within the category. It is very unlikely that either of us gotthe image in our mind’s eye of a mouse larger than an elephant stridingover an elephant smaller than a mouse. Instead, we each separately butjointly visualized a mouse smaller than a shoe clambering over an elephantlarger than a sofa. System 1, which understands language, has access tonorms of categories, which specify the range of plausible values as well asthe most typical cases.

Seeing Causes and Intentions

“Fred’s parents arrived late. The caterers were expected soon. Fred wasangry.” You know why Fred was angry, and it is not because the catererswere expected soon. In your network of associationsmals in co, anger andlack of punctuality are linked as an effect and its possible cause, but thereis no such link between anger and the idea of expecting caterers. Acoherent story was instantly constructed as you read; you immediatelyknew the cause of Fred’s anger. Finding such causal connections is part ofunderstanding a story and is an automatic operation of System 1. System2, your conscious self, was offered the causal interpretation and acceptedit.

A story in Nassim Taleb’s The Black Swan illustrates this automaticsearch for causality. He reports that bond prices initially rose on the day ofSaddam Hussein’s capture in his hiding place in Iraq. Investors wereapparently seeking safer assets that morning, and the Bloomberg Newsservice flashed this headline: U.S. TREASURIES RISE; HUSSEIN CAPTURE MAY NOTCURB TERRORISM. Half an hour later, bond prices fell back and the revisedheadline read: U.S. TREASURIES FALL; HUSSEIN CAPTURE BOOSTS ALLURE OFRISKY ASSETS. Obviously, Hussein’s capture was the major event of the day,and because of the way the automatic search for causes shapes ourthinking, that event was destined to be the explanation of whateverhappened in the market on that day. The two headlines look superficially

(PDF) Thinking, fast and slow by daniel kahneman - PDFSLIDE.US (76)

like explanations of what happened in the market, but a statement that canexplain two contradictory outcomes explains nothing at all. In fact, all theheadlines do is satisfy our need for coherence: a large event is supposedto have consequences, and consequences need causes to explain them.We have limited information about what happened on a day, and System 1is adept at finding a coherent causal story that links the fragments ofknowledge at its disposal.

Read this sentence:

After spending a day exploring beautiful sights in the crowdedstreets of New York, Jane discovered that her wallet was missing.

When people who had read this brief story (along with many others) weregiven a surprise recall test, the word pickpocket was more stronglyassociated with the story than the word sights, even though the latter wasactually in the sentence while the former was not. The rules of associativecoherence tell us what happened. The event of a lost wallet could evokemany different causes: the wallet slipped out of a pocket, was left in therestaurant, etc. However, when the ideas of lost wallet, New York, andcrowds are juxtaposed, they jointly evoke the explanation that a pickpocketcaused the loss. In the story of the startling soup, the outcome—whetheranother customer wincing at the taste of the soup or the first person’sextreme reaction to the waiter’s touch—brings about an associativelycoherent interpretation of the initial surprise, completing a plausible story.

The aristocratic Belgian psychologist Albert Michotte published a bookin 1945 (translated into English in 1963) that overturned centuries ofthinking about causality, going back at least to Hume’s examination of theassociation of ideas. The commonly accepted wisdom was that we inferphysical causality from repeated observations of correlations amongevents. We have had myriad experiences in which we saw one object inmotion touching another object, which immediately starts to move, often(but not always) in the same direction. This is what happens when a billiardball hits another, and it is also what happens when you knock over a vaseby brushing against it. Michotte had a different idea: he argued that we seecausality, just as directly as we see color. To make his point, he createdepisodes in n ttiowhich a black square drawn on paper is seen in motion; itcomes into contact with another square, which immediately begins tomove. The observers know that there is no real physical contact, but theynevertheless have a powerful “illusion of causality.” If the second objectstarts moving instantly, they describe it as having been “launched” by thefirst. Experiments have shown that six-month-old infants see the sequenceof events as a cause-effect scenario, and they indicate surprise when the

(PDF) Thinking, fast and slow by daniel kahneman - PDFSLIDE.US (77)

sequence is altered. We are evidently ready from birth to haveimpressions of causality, which do not depend on reasoning aboutpatterns of causation. They are products of System 1.

In 1944, at about the same time as Michotte published hisdemonstrations of physical causality, the psychologists Fritz Heider andMary-Ann Simmel used a method similar to Michotte’s to demonstrate theperception of intentional causality. They made a film, which lasts all of oneminute and forty seconds, in which you see a large triangle, a smalltriangle, and a circle moving around a shape that looks like a schematicview of a house with an open door. Viewers see an aggressive largetriangle bullying a smaller triangle, a terrified circle, the circle and the smalltriangle joining forces to defeat the bully; they also observe muchinteraction around a door and then an explosive finale. The perception ofintention and emotion is irresistible; only people afflicted by autism do notexperience it. All this is entirely in your mind, of course. Your mind is readyand even eager to identify agents, assign them personality traits andspecific intentions, and view their actions as expressing individualpropensities. Here again, the evidence is that we are born prepared tomake intentional attributions: infants under one year old identify bullies andvictims, and expect a pursuer to follow the most direct path in attempting tocatch whatever it is chasing.

The experience of freely willed action is quite separate from physicalcausality. Although it is your hand that picks up the salt, you do not think ofthe event in terms of a chain of physical causation. You experience it ascaused by a decision that a disembodied you made, because you wantedto add salt to your food. Many people find it natural to describe their soulas the source and the cause of their actions. The psychologist Paul Bloom,writing in The Atlantic in 2005, presented the provocative claim that ourinborn readiness to separate physical and intentional causality explains thenear universality of religious beliefs. He observes that “we perceive theworld of objects as essentially separate from the world of minds, making itpossible for us to envision soulless bodies and bodiless souls.” The twomodes of causation that we are set to perceive make it natural for us toaccept the two central beliefs of many religions: an immaterial divinity isthe ultimate cause of the physical world, and immortal souls temporarilycontrol our bodies while we live and leave them behind as we die. InBloom’s view, the two concepts of causality were shaped separately byevolutionary forces, building the origins of religion into the structure ofSystem 1.

The prominence of causal intuitions is a recurrent theme in this bookbecause people are prone to apply causal thinking inappropriately, to

(PDF) Thinking, fast and slow by daniel kahneman - PDFSLIDE.US (78)

situations that require statistical reasoning. Statistical thinking derivesconclusions about individual cases from properties of categories andensembles. Unfortunately, System 1 does not have the capability for thismode of reasoning; System 2 can learn to think statistically, but few peoplereceive the necessary training.

The psychology of causality was the basis of my decision to describepsycl c to thinhological processes by metaphors of agency, with littleconcern for consistency. I sometimes refer to System 1 as an agent withcertain traits and preferences, and sometimes as an associative machinethat represents reality by a complex pattern of links. The system and themachine are fictions; my reason for using them is that they fit the way wethink about causes. Heider’s triangles and circles are not really agents—itis just very easy and natural to think of them that way. It is a matter ofmental economy. I assume that you (like me) find it easier to think aboutthe mind if we describe what happens in terms of traits and intentions (thetwo systems) and sometimes in terms of mechanical regularities (theassociative machine). I do not intend to convince you that the systems arereal, any more than Heider intended you to believe that the large triangle isreally a bully.

Speaking of Norms and Causes

“When the second applicant also turned out to be an old friend ofmine, I wasn’t quite as surprised. Very little repetition is neededfor a new experience to feel normal!”

“When we survey the reaction to these products, let’s make surewe don’t focus exclusively on the average. We should considerthe entire range of normal reactions.”

“She can’t accept that she was just unlucky; she needs a causalstory. She will end up thinking that someone intentionallysabotaged her work.”

(PDF) Thinking, fast and slow by daniel kahneman - PDFSLIDE.US (79)

A Machine for Jumping to Conclusions

The great comedian Danny Kaye had a line that has stayed with me sincemy adolescence. Speaking of a woman he dislikes, he says, “Her favoriteposition is beside herself, and her favorite sport is jumping to conclusions.”The line came up, I remember, in the initial conversation with AmosTversky about the rationality of statistical intuitions, and now I believe itoffers an apt description of how System 1 functions. Jumping toconclusions is efficient if the conclusions are likely to be correct and thecosts of an occasional mistake acceptable, and if the jump saves muchtime and effort. Jumping to conclusions is risky when the situation isunfamiliar, the stakes are high, and there is no time to collect moreinformation. These are the circumstances in which intuitive errors areprobable, which may be prevented by a deliberate intervention of System2.

Neglect of Ambiguity and Suppression of Doubt

Figure 6

What do the three exhibits in figure 6 have in common? The answer is thatall are ambiguous. You almost certainly read the display on the left as A BC and the one on the right as 12 13 14, but the middle items in bothdisplays are identical. You could just as well have read e iom prthe cvethem as A 13 C or 12 B 14, but you did not. Why not? The same shape isread as a letter in a context of letters and as a number in a context ofnumbers. The entire context helps determine the interpretation of eachelement. The shape is ambiguous, but you jump to a conclusion about itsidentity and do not become aware of the ambiguity that was resolved.

As for Ann, you probably imagined a woman with money on her mind,walking toward a building with tellers and secure vaults. But this plausibleinterpretation is not the only possible one; the sentence is ambiguous. If anearlier sentence had been “They were floating gently down the river,” youwould have imagined an altogether different scene. When you have justbeen thinking of a river, the word bank is not associated with money. In the

(PDF) Thinking, fast and slow by daniel kahneman - PDFSLIDE.US (80)

absence of an explicit context, System 1 generated a likely context on itsown. We know that it is System 1 because you were not aware of thechoice or of the possibility of another interpretation. Unless you have beencanoeing recently, you probably spend more time going to banks thanfloating on rivers, and you resolved the ambiguity accordingly. Whenuncertain, System 1 bets on an answer, and the bets are guided byexperience. The rules of the betting are intelligent: recent events and thecurrent context have the most weight in determining an interpretation.When no recent event comes to mind, more distant memories govern.Among your earliest and most memorable experiences was singing yourABCs; you did not sing your A13Cs.

The most important aspect of both examples is that a definite choicewas made, but you did not know it. Only one interpretation came to mind,and you were never aware of the ambiguity. System 1 does not keep trackof alternatives that it rejects, or even of the fact that there were alternatives.Conscious doubt is not in the repertoire of System 1; it requiresmaintaining incompatible interpretations in mind at the same time, whichdemands mental effort. Uncertainty and doubt are the domain of System 2.

A Bias to Believe and Confirm

The psychologist Daniel Gilbert, widely known as the author of Stumblingto Happiness, once wrote an essay, titled “How Mental Systems Believe,”in which he developed a theory of believing and unbelieving that he tracedto the seventeenth-century philosopher Baruch Spinoza. Gilbert proposedthat understanding a statement must begin with an attempt to believe it:you must first know what the idea would mean if it were true. Only then canyou decide whether or not to unbelieve it. The initial attempt to believe isan automatic operation of System 1, which involves the construction of thebest possible interpretation of the situation. Even a nonsensical statement,Gilbert argues, will evoke initial belief. Try his example: “whitefish eatcandy.” You probably were aware of vague impressions of fish and candyas an automatic process of associative memory searched for linksbetween the two ideas that would make sense of the nonsense.

Gilbert sees unbelieving as an operation of System 2, and he reportedan elegant experiment to make his point. The participants saw nonsensicalassertions, such as “a dinca is a flame,” followed after a few seconds by asingle word, “true” or “false.” They were later tested for their memory ofwhich sentences had been labeled “true.” In one condition of theexperiment subjects were required to hold digits in memory during thetask. The disruption of System 2 had a selective effect: it made it difficult

(PDF) Thinking, fast and slow by daniel kahneman - PDFSLIDE.US (81)

for people to “unbelieve” false sentences. In a later test of memory, thedepleted par muumbling toticipants ended up thinking that many of thefalse sentences were true. The moral is significant: when System 2 isotherwise engaged, we will believe almost anything. System 1 is gullibleand biased to believe, System 2 is in charge of doubting and unbelieving,but System 2 is sometimes busy, and often lazy. Indeed, there is evidencethat people are more likely to be influenced by empty persuasivemessages, such as commercials, when they are tired and depleted.

The operations of associative memory contribute to a generalconfirmation bias. When asked, “Is Sam friendly?” different instances ofSam’s behavior will come to mind than would if you had been asked “IsSam unfriendly?” A deliberate search for confirming evidence, known aspositive test strategy, is also how System 2 tests a hypothesis. Contrary tothe rules of philosophers of science, who advise testing hypotheses bytrying to refute them, people (and scientists, quite often) seek data that arelikely to be compatible with the beliefs they currently hold. The confirmatorybias of System 1 favors uncritical acceptance of suggestions andexaggeration of the likelihood of extreme and improbable events. If you areasked about the probability of a tsunami hitting California within the nextthirty years, the images that come to your mind are likely to be images oftsunamis, in the manner Gilbert proposed for nonsense statements suchas “whitefish eat candy.” You will be prone to overestimate the probabilityof a disaster.

Exaggerated Emotional Coherence (Halo Effect)

If you like the president’s politics, you probably like his voice and hisappearance as well. The tendency to like (or dislike) everything about aperson—including things you have not observed—is known as the haloeffect. The term has been in use in psychology for a century, but it has notcome into wide use in everyday language. This is a pity, because the haloeffect is a good name for a common bias that plays a large role in shapingour view of people and situations. It is one of the ways the representationof the world that System 1 generates is simpler and more coherent thanthe real thing.

You meet a woman named Joan at a party and find her personable andeasy to talk to. Now her name comes up as someone who could be askedto contribute to a charity. What do you know about Joan’s generosity? Thecorrect answer is that you know virtually nothing, because there is littlereason to believe that people who are agreeable in social situations arealso generous contributors to charities. But you like Joan and you will

(PDF) Thinking, fast and slow by daniel kahneman - PDFSLIDE.US (82)

retrieve the feeling of liking her when you think of her. You also likegenerosity and generous people. By association, you are nowpredisposed to believe that Joan is generous. And now that you believeshe is generous, you probably like Joan even better than you did earlier,because you have added generosity to her pleasant attributes.

Real evidence of generosity is missing in the story of Joan, and the gapis filled by a guess that fits one’s emotional response to her. In othersituations, evidence accumulates gradually and the interpretation isshaped by the emotion attached to the first impression. In an enduringclassic of psychology, Solomon Asch presented descriptions of twopeople and asked for comments on their personality. What do you think ofAlan and Ben?

Alan: intelligent—industrious—impulsive—critical—stubborn—enviousBen: envious—The#82stubborn—critical—impulsive—industrious—intelligent

If you are like most of us, you viewed Alan much more favorably than Ben.The initial traits in the list change the very meaning of the traits that appearlater. The stubbornness of an intelligent person is seen as likely to bejustified and may actually evoke respect, but intelligence in an envious andstubborn person makes him more dangerous. The halo effect is also anexample of suppressed ambiguity: like the word bank, the adjectivestubborn is ambiguous and will be interpreted in a way that makes itcoherent with the context.

There have been many variations on this research theme. Participants inone study first considered the first three adjectives that describe Alan; thenthey considered the last three, which belonged, they were told, to anotherperson. When they had imagined the two individuals, the participants wereasked if it was plausible for all six adjectives to describe the same person,and most of them thought it was impossible!

The sequence in which we observe characteristics of a person is oftendetermined by chance. Sequence matters, however, because the haloeffect increases the weight of first impressions, sometimes to the point thatsubsequent information is mostly wasted. Early in my career as aprofessor, I graded students’ essay exams in the conventional way. I wouldpick up one test booklet at a time and read all that student’s essays inimmediate succession, grading them as I went. I would then compute thetotal and go on to the next student. I eventually noticed that my evaluationsof the essays in each booklet were strikingly homogeneous. I began tosuspect that my grading exhibited a halo effect, and that the first question I

(PDF) Thinking, fast and slow by daniel kahneman - PDFSLIDE.US (83)

scored had a disproportionate effect on the overall grade. The mechanismwas simple: if I had given a high score to the first essay, I gave the studentthe benefit of the doubt whenever I encountered a vague or ambiguousstatement later on. This seemed reasonable. Surely a student who haddone so well on the first essay would not make a foolish mistake in thesecond one! But there was a serious problem with my way of doing things.If a student had written two essays, one strong and one weak, I would endup with different final grades depending on which essay I read first. I hadtold the students that the two essays had equal weight, but that was nottrue: the first one had a much greater impact on the final grade than thesecond. This was unacceptable.

I adopted a new procedure. Instead of reading the booklets in sequence,I read and scored all the students’ answers to the first question, then wenton to the next one. I made sure to write all the scores on the inside backpage of the booklet so that I would not be biased (even unconsciously)when I read the second essay. Soon after switching to the new method, Imade a disconcerting observation: my confidence in my grading was nowmuch lower than it had been. The reason was that I frequently experienceda discomfort that was new to me. When I was disappointed with astudent’s second essay and went to the back page of the booklet to entera poor grade, I occasionally discovered that I had given a top grade to thesame student’s first essay. I also noticed that I was tempted to reduce thediscrepancy by changing the grade that I had not yet written down, andfound it hard to follow the simple rule of never yielding to that temptation.My grades for the essays of a single student often varied over aconsiderable range. The lack of coherence left me uncertain andfrustrated.

I was now less happy with and less confident in my grades than I hadbeen earlier, but I recognized that thass confthis was a good sign, anindication that the new procedure was superior. The consistency I hadenjoyed earlier was spurious; it produced a feeling of cognitive ease, andmy System 2 was happy to lazily accept the final grade. By allowing myselfto be strongly influenced by the first question in evaluating subsequentones, I spared myself the dissonance of finding the same student doingvery well on some questions and badly on others. The uncomfortableinconsistency that was revealed when I switched to the new procedure wasreal: it reflected both the inadequacy of any single question as a measureof what the student knew and the unreliability of my own grading.

The procedure I adopted to tame the halo effect conforms to a generalprinciple: decorrelate error! To understand how this principle works,imagine that a large number of observers are shown glass jars containingpennies and are challenged to estimate the number of pennies in each jar.

(PDF) Thinking, fast and slow by daniel kahneman - PDFSLIDE.US (84)

pennies and are challenged to estimate the number of pennies in each jar.As James Surowiecki explained in his best-selling The Wisdom ofCrowds, this is the kind of task in which individuals do very poorly, butpools of individual judgments do remarkably well. Some individuals greatlyoverestimate the true number, others underestimate it, but when manyjudgments are averaged, the average tends to be quite accurate. Themechanism is straightforward: all individuals look at the same jar, and alltheir judgments have a common basis. On the other hand, the errors thatindividuals make are independent of the errors made by others, and (in theabsence of a systematic bias) they tend to average to zero. However, themagic of error reduction works well only when the observations areindependent and their errors uncorrelated. If the observers share a bias,the aggregation of judgments will not reduce it. Allowing the observers toinfluence each other effectively reduces the size of the sample, and with itthe precision of the group estimate.

To derive the most useful information from multiple sources of evidence,you should always try to make these sources independent of each other.This rule is part of good police procedure. When there are multiplewitnesses to an event, they are not allowed to discuss it before giving theirtestimony. The goal is not only to prevent collusion by hostile witnesses, itis also to prevent unbiased witnesses from influencing each other.Witnesses who exchange their experiences will tend to make similar errorsin their testimony, reducing the total value of the information they provide.Eliminating redundancy from your sources of information is always a goodidea.

The principle of independent judgments (and decorrelated errors) hasimmediate applications for the conduct of meetings, an activity in whichexecutives in organizations spend a great deal of their working days. Asimple rule can help: before an issue is discussed, all members of thecommittee should be asked to write a very brief summary of their position.This procedure makes good use of the value of the diversity of knowledgeand opinion in the group. The standard practice of open discussion givestoo much weight to the opinions of those who speak early and assertively,causing others to line up behind them.

What You See is All There is (Wysiati)

One of my favorite memories of the early years of working with Amos is acomedy routine he enjoyed performing. In a perfect impersonation of oneof the professors with whom he had studied philosophy as anundergraduate, Amos would growl in Hebrew marked by a thick Germanaccent: “You must never forget the Primat of the Is.” What exactly his

(PDF) Thinking, fast and slow by daniel kahneman - PDFSLIDE.US (85)

teacher had meant by that phrase never became clear to me (or to Amos, Ibelieve), but Amos’s jokes always maht=cipde a point. He was remindedof the old phrase (and eventually I was too) whenever we encountered theremarkable asymmetry between the ways our mind treats information thatis currently available and information we do not have.

An essential design feature of the associative machine is that itrepresents only activated ideas. Information that is not retrieved (evenunconsciously) from memory might as well not exist. System 1 excels atconstructing the best possible story that incorporates ideas currentlyactivated, but it does not (cannot) allow for information it does not have.

The measure of success for System 1 is the coherence of the story itmanages to create. The amount and quality of the data on which the storyis based are largely irrelevant. When information is scarce, which is acommon occurrence, System 1 operates as a machine for jumping toconclusions. Consider the following: “Will Mindik be a good leader? She isintelligent and strong…” An answer quickly came to your mind, and it wasyes. You picked the best answer based on the very limited informationavailable, but you jumped the gun. What if the next two adjectives werecorrupt and cruel?

Take note of what you did not do as you briefly thought of Mindik as aleader. You did not start by asking, “What would I need to know before Iformed an opinion about the quality of someone’s leadership?” System 1got to work on its own from the first adjective: intelligent is good, intelligentand strong is very good. This is the best story that can be constructed fromtwo adjectives, and System 1 delivered it with great cognitive ease. Thestory will be revised if new information comes in (such as Mindik iscorrupt), but there is no waiting and no subjective discomfort. And therealso remains a bias favoring the first impression.

The combination of a coherence-seeking System 1 with a lazy System 2implies that System 2 will endorse many intuitive beliefs, which closelyreflect the impressions generated by System 1. Of course, System 2 alsois capable of a more systematic and careful approach to evidence, and offollowing a list of boxes that must be checked before making a decision—think of buying a home, when you deliberately seek information that youdon’t have. However, System 1 is expected to influence even the morecareful decisions. Its input never ceases.

Jumping to conclusions on the basis of limited evidence is so importantto an understanding of intuitive thinking, and comes up so often in thisbook, that I will use a cumbersome abbreviation for it: WYSIATI, whichstands for what you see is all there is. System 1 is radically insensitive toboth the quality and the quantity of the information that gives rise to

(PDF) Thinking, fast and slow by daniel kahneman - PDFSLIDE.US (86)

impressions and intuitions.Amos, with two of his graduate students at Stanford, reported a study

that bears directly on WYSIATI, by observing the reaction of people whoare given one-sided evidence and know it. The participants were exposedto legal scenarios such as the following:

On September 3, plaintiff David Thornton, a forty-three-year-oldunion field representative, was present in Thrifty Drug Store#168, performing a routine union visit. Within ten minutes of hisarrival, a store manager confronted him and told him he could nolonger speak with the union employees on the floor of the store.Instead, he would have to see them in a back room while theywere on break. Such a request is allowed by the union contractwith Thrifty Drug but had never before been enforced. When Mr.Thornton objected, he was told that he had the choice of contoroom whilforming to these requirements, leaving the store, orbeing arrested. At this point, Mr. Thornton indicated to themanager that he had always been allowed to speak toemployees on the floor for as much as ten minutes, as long as nobusiness was disrupted, and that he would rather be arrestedthan change the procedure of his routine visit. The manager thencalled the police and had Mr. Thornton handcuffed in the store fortrespassing. After he was booked and put into a holding cell for abrief time, all charges were dropped. Mr. Thornton is suing ThriftyDrug for false arrest.

In addition to this background material, which all participants read, differentgroups were exposed to presentations by the lawyers for the two parties.Naturally, the lawyer for the union organizer described the arrest as anintimidation attempt, while the lawyer for the store argued that having thetalk in the store was disruptive and that the manager was acting properly.Some participants, like a jury, heard both sides. The lawyers added nouseful information that you could not infer from the background story.

The participants were fully aware of the setup, and those who heard onlyone side could easily have generated the argument for the other side.Nevertheless, the presentation of one-sided evidence had a verypronounced effect on judgments. Furthermore, participants who saw one-sided evidence were more confident of their judgments than those whosaw both sides. This is just what you would expect if the confidence thatpeople experience is determined by the coherence of the story theymanage to construct from available information. It is the consistency of theinformation that matters for a good story, not its completeness. Indeed, you

(PDF) Thinking, fast and slow by daniel kahneman - PDFSLIDE.US (87)

will often find that knowing little makes it easier to fit everything you knowinto a coherent pattern.

WY SIATI facilitates the achievement of coherence and of the cognitiveease that causes us to accept a statement as true. It explains why we canthink fast, and how we are able to make sense of partial information in acomplex world. Much of the time, the coherent story we put together isclose enough to reality to support reasonable action. However, I will alsoinvoke WY SIATI to help explain a long and diverse list of biases ofjudgment and choice, including the following among many others:

Overconfidence: As the WY SIATI rule implies, neither the quantitynor the quality of the evidence counts for much in subjectiveconfidence. The confidence that individuals have in their beliefsdepends mostly on the quality of the story they can tell about whatthey see, even if they see little. We often fail to allow for thepossibility that evidence that should be critical to our judgment ismissing—what we see is all there is. Furthermore, our associativesystem tends to settle on a coherent pattern of activation andsuppresses doubt and ambiguity.Framing effects: Different ways of presenting the same informationoften evoke different emotions. The statement that “the odds ofsurvival one month after surgery are 90%” is more reassuring thanthe equivalent statement that “mortality within one month of surgery is10%.” Similarly, cold cuts described as “90% fat-free” are moreattractive than when they are described as “10% fat.” Theequivalence of the alternative formulations is transparent, but anindividual normally sees only one formulation, and what she sees isall there is.Base-rate neglect: Recall Steve, the meek and tidy soul who is oftenbelieved to be a librarian. The personality description is salient andvivid, and although you surely know that there are more male farm muBase-rers than male librarians, that statistical fact almost certainlydid not come to your mind when you first considered the question.What you saw was all there was.

Speaking of Jumping to Conclusions

(PDF) Thinking, fast and slow by daniel kahneman - PDFSLIDE.US (88)

“She knows nothing about this person’s management skills. Allshe is going by is the halo effect from a good presentation.”

“Let’s decorrelate errors by obtaining separate judgments on theissue before any discussion. We will get more information fromindependent assessments.”

“They made that big decision on the basis of a good report fromone consultant. WYSIATI—what you see is all there is. They didnot seem to realize how little information they had.”

“They didn’t want more information that might spoil their story.WYSIATI.”

(PDF) Thinking, fast and slow by daniel kahneman - PDFSLIDE.US (89)

How Judgments Happen

There is no limit to the number of questions you can answer, whether theyare questions someone else asks or questions you ask yourself. Nor isthere a limit to the number of attributes you can evaluate. You are capableof counting the number of capital letters on this page, comparing the heightof the windows of your house to the one across the street, and assessingthe political prospects of your senator on a scale from excellent todisastrous. The questions are addressed to System 2, which will directattention and search memory to find the answers. System 2 receivesquestions or generates them: in either case it directs attention andsearches memory to find the answers. System 1 operates differently. Itcontinuously monitors what is going on outside and inside the mind, andcontinuously generates assessments of various aspects of the situationwithout specific intention and with little or no effort. These basicassessments play an important role in intuitive judgment, because they areeasily substituted for more difficult questions—this is the essential idea ofthe heuristics and biases approach. Two other features of System 1 alsosupport the substitution of one judgment for another. One is the ability totranslate values across dimensions, which you do in answering a questionthat most people find easy: “If Sam were as tall as he is intelligent, how tallwould he be?” Finally, there is the mental shotgun. An intention of System 2to answer a specific question or evaluate a particular attribute of thesituation automatically triggers other computations, including basicassessments.

Basic Assessments

System 1 has been shaped by evolution to provide a continuousassessment of the main problems that an organism must solve to survive:How are things going? Is there a threat or a major opportunity? Iseverything normal? Should I approach or avoid? The questions areperhaps less urgent for a human in a city environment than for a gazelle onthe savannah, aalenc and e: How , but we have inherited the neuralmechanisms that evolved to provide ongoing assessments of threat level,and they have not been turned off. Situations are constantly evaluated asgood or bad, requiring escape or permitting approach. Good mood andcognitive ease are the human equivalents of assessments of safety andfamiliarity.

For a specific example of a basic assessment, consider the ability todiscriminate friend from foe at a glance. This contributes to one’s chances

(PDF) Thinking, fast and slow by daniel kahneman - PDFSLIDE.US (90)

of survival in a dangerous world, and such a specialized capability hasindeed evolved. Alex Todorov, my colleague at Princeton, has explored thebiological roots of the rapid judgments of how safe it is to interact with astranger. He showed that we are endowed with an ability to evaluate, in asingle glance at a stranger’s face, two potentially crucial facts about thatperson: how dominant (and therefore potentially threatening) he is, andhow trustworthy he is, whether his intentions are more likely to be friendly orhostile. The shape of the face provides the cues for assessing dominance:a “strong” square chin is one such cue. Facial expression (smile or frown)provides the cues for assessing the stranger’s intentions. The combinationof a square chin with a turned-down mouth may spell trouble. The accuracyof face reading is far from perfect: round chins are not a reliable indicatorof meekness, and smiles can (to some extent) be faked. Still, even animperfect ability to assess strangers confers a survival advantage.

This ancient mechanism is put to a novel use in the modern world: it hassome influence on how people vote. Todorov showed his students picturesof men’s faces, sometimes for as little as one-tenth of a second, andasked them to rate the faces on various attributes, including likability andcompetence. Observers agreed quite well on those ratings. The faces thatTodorov showed were not a random set: they were the campaign portraitsof politicians competing for elective office. Todorov then compared theresults of the electoral races to the ratings of competence that Princetonstudents had made, based on brief exposure to photographs and withoutany political context. In about 70% of the races for senator, congressman,and governor, the election winner was the candidate whose face hadearned a higher rating of competence. This striking result was quicklyconfirmed in national elections in Finland, in zoning board elections inEngland, and in various electoral contests in Australia, Germany, andMexico. Surprisingly (at least to me), ratings of competence were far morepredictive of voting outcomes in Todorov’s study than ratings of likability.

Todorov has found that people judge competence by combining the twodimensions of strength and trustworthiness. The faces that exudecompetence combine a strong chin with a slight confident-appearingsmile. There is no evidence that these facial features actually predict howwell politicians will perform in office. But studies of the brain’s response towinning and losing candidates show that we are biologically predisposedto reject candidates who lack the attributes we value—in this research,losers evoked stronger indications of (negative) emotional response. Thisis an example of what I will call a judgment heuristic in the followingchapters. Voters are attempting to form an impression of how good acandidate will be in office, and they fall back on a simpler assessment thatis made quickly and automatically and is available when System 2 must

(PDF) Thinking, fast and slow by daniel kahneman - PDFSLIDE.US (91)

is made quickly and automatically and is available when System 2 mustmake its decision.

Political scientists followed up on Todorov’s initial research byidentifying a category of voters for whom the automatic preferences ofSystem 1 are particularly likely to play a large role. They found what theywere looking for among politicalr m="5%">Todoly uninformed voters whowatch a great deal of television. As expected, the effect of facialcompetence on voting is about three times larger for information-poor andTV-prone voters than for others who are better informed and watch lesstelevision. Evidently, the relative importance of System 1 in determiningvoting choices is not the same for all people. We will encounter otherexamples of such individual differences.

System 1 understands language, of course, and understanding dependson the basic assessments that are routinely carried out as part of theperception of events and the comprehension of messages. Theseassessments include computations of similarity and representativeness,attributions of causality, and evaluations of the availability of associationsand exemplars. They are performed even in the absence of a specific taskset, although the results are used to meet task demands as they arise.

The list of basic assessments is long, but not every possible attribute isassessed. For an example, look briefly at figure 7.

A glance provides an immediate impression of many features of thedisplay. You know that the two towers are equally tall and that they aremore similar to each other than the tower on the left is to the array of blocksin the middle. However, you do not immediately know that the number ofblocks in the left-hand tower is the same as the number of blocks arrayedon the floor, and you have no impression of the height of the tower that youcould build from them. To confirm that the numbers are the same, youwould need to count the two sets of blocks and compare the results, anactivity that only System 2 can carry out.

(PDF) Thinking, fast and slow by daniel kahneman - PDFSLIDE.US (92)

Figure 7

Sets and Prototypes

For another example, consider the question: What is the average length ofthe lines in figure 8?

Figure 8

This question is easy and System 1 answers it without prompting.Experiments have shown that a fraction of a second is sufficient for peopleto register the average length of an array of lines with considerableprecision. Furthermore, the accuracy of these judgments is not impairedwhen the observer is cognitively busy with a memory task. They do notnecessarily know how to describe the average in inches or centimeters,but they will be very accurate in adjusting the length of another line to matchthe average. System 2 is not needed to form an impression of the norm oflength for an array. System 1 does it, automatically and effortlessly, just asit registers the color of the lines and the fact that they are not parallel. Wealso can form an immediate impression of the number of objects in anarray—precisely if there are four or fewer objects, crudely if there aremore.

Now to another question: What is the total length of the lines in figure 8?This is a different experience, because System 1 has no suggestions tooffer. The only way you can answer this question is by activating System 2,which will laboriously estimate the average, estimate or count the lines,and multiply average length by the number of lines.estimaight="0%">

The failure of System 1 to compute the total length of a set of lines at aglance may look obvious to you; you never thought you could do it. It is infact an instance of an important limitation of that system. Because System1 represents categories by a prototype or a set of typical exemplars, it

(PDF) Thinking, fast and slow by daniel kahneman - PDFSLIDE.US (93)

deals well with averages but poorly with sums. The size of the category, thenumber of instances it contains, tends to be ignored in judgments of what Iwill call sum-like variables.

Participants in one of the numerous experiments that were prompted bythe litigation following the disastrous Exxon Valdez oil spill were askedtheir willingness to pay for nets to cover oil ponds in which migratory birdsoften drown. Different groups of participants stated their willingness to payto save 2,000, 20,000, or 200,000 birds. If saving birds is an economicgood it should be a sum-like variable: saving 200,000 birds should beworth much more than saving 2,000 birds. In fact, the average contributionsof the three groups were $80, $78, and $88 respectively. The number ofbirds made very little difference. What the participants reacted to, in allthree groups, was a prototype—the awful image of a helpless birddrowning, its feathers soaked in thick oil. The almost complete neglect ofquantity in such emotional contexts has been confirmed many times.

Intensity Matching

Questions about your happiness, the president’s popularity, the properpunishment of financial evildoers, and the future prospects of a politicianshare an important characteristic: they all refer to an underlying dimensionof intensity or amount, which permits the use of the word more: morehappy, more popular, more severe, or more powerful (for a politician). Forexample, a candidate’s political future can range from the low of “She willbe defeated in the primary” to a high of “She will someday be president ofthe United States.”

Here we encounter a new aptitude of System 1. An underlying scale ofintensity allows matching across diverse dimensions. If crimes werecolors, murder would be a deeper shade of red than theft. If crimes wereexpressed as music, mass murder would be played fortissimo whileaccumulating unpaid parking tickets would be a faint pianissimo. And ofcourse you have similar feelings about the intensity of punishments. Inclassic experiments, people adjusted the loudness of a sound to theseverity of crimes; other people adjusted loudness to the severity of legalpunishments. If you heard two notes, one for the crime and one for thepunishment, you would feel a sense of injustice if one tone was muchlouder than the other.

Consider an example that we will encounter again later:

Julie read fluently when she was four years old.

(PDF) Thinking, fast and slow by daniel kahneman - PDFSLIDE.US (94)

Now match Julie’s reading prowess as a child to the following intensityscales:

How tall is a man who is as tall as Julie was precocious?

What do you think of 6 feet? Obviously too little. What about 7 feet?Probably too much. You are looking for a height that is as remarkable asthe achievement of reading at age four. Fairly remarkable, but notextraordinary. Reading at fifteen months would be extraordinary, perhapslike a man who is 7'8".

What level of income in your profession matches Julie’s readingachievement?Which crime is as severe as Julie was precocious?Which graduating GPA in an Ivy League college matches Julie’sreading?

Not very hard, was it? Furthermore, you can be assured that your matcheswill be quite close to those of other people in your cultural milieu. We willsee that when people are asked to predict Julie’s GPA from theinformation about the age at which she learned to read, they answer bytranslating from one scale to another and pick the matching GPA. And wewill also see why this mode of prediction by matching is statistically wrong—although it is perfectly natural to System 1, and for most people exceptstatisticians it is also acceptable to System 2.

The Mental Shotgun

System 1 carries out many computations at any one time. Some of theseare routine assessments that go on continuously. Whenever your eyes areopen, your brain computes a three-dimensional representation of what is inyour field of vision, complete with the shape of objects, their position inspace, and their identity. No intention is needed to trigger this operation orthe continuous monitoring for violated expectations. In contrast to theseroutine assessments, other computations are undertaken only whenneeded: you do not maintain a continuous evaluation of how happy orwealthy you are, and even if you are a political addict you do notcontinuously assess the president’s prospects. The occasional judgmentsare voluntary. They occur only when you intend them to do so.

You do not automatically count the number of syllables of every word youread, but you can do it if you so choose. However, the control overintended computations is far from precise: we often compute much morethan we want or need. I call this excess computation the mental shotgun. It

(PDF) Thinking, fast and slow by daniel kahneman - PDFSLIDE.US (95)

than we want or need. I call this excess computation the mental shotgun. Itis impossible to aim at a single point with a shotgun because it shootspellets that scatter, and it seems almost equally difficult for System 1 not todo more than System 2 charges it to do. Two experiments that I read longago suggested this image.

Participants in one experiment listened to pairs of words, with theinstruction to press a key as quickly as possible whenever they detectedthat the words rhymed. The words rhyme in both these pairs:

VOTE—NOTEVOTE—GOAT

The difference is obvious to you because you see the two pairs. VOTE andGOAT rhyme, but they are spelled differently. The participants only heardthe words, but they were also influenced by the spelling. They weredistinctly slower to recognize the words as rhyming if their spelling wasdiscrepant. Although the instructions required only a comparison ofsounds, the participants also compared their spelling, and the mismatchon the irrelevant dimension slowed them down. An intention to answer onequestion evoked another, which was not only superfluous but actuallydetrimental to the main task.

In another study, people listened to a series of sentences, with theinstruction to press one key as quickly as post="lly desible to indicate if thesentence was literally true, and another key if the sentence was not literallytrue. What are the correct responses for the following sentences?

Some roads are snakes.Some jobs are snakes.Some jobs are jails.

All three sentences are literally false. However, you probably noticed thatthe second sentence is more obviously false than the other two—thereaction times collected in the experiment confirmed a substantialdifference. The reason for the difference is that the two difficult sentencescan be metaphorically true. Here again, the intention to perform onecomputation evoked another. And here again, the correct answer prevailedin the conflict, but the conflict with the irrelevant answer disruptedperformance. In the next chapter we will see that the combination of amental shotgun with intensity matching explains why we have intuitivejudgments about many things that we know little about.

Speaking of Judgment

(PDF) Thinking, fast and slow by daniel kahneman - PDFSLIDE.US (96)

“Evaluating people as attractive or not is a basic assessment.You do that automatically whether or not you want to, and itinfluences you.”

“There are circuits in the brain that evaluate dominance from theshape of the face. He looks the part for a leadership role.”

“The punishment won’t feel just unless its intensity matches thecrime. Just like you can match the loudness of a sound to thebrightness of a light.”

“This was a clear instance of a mental shotgun. He was askedwhether he thought the company was financially sound, but hecouldn’t forget that he likes their product.”

(PDF) Thinking, fast and slow by daniel kahneman - PDFSLIDE.US (97)

Answering an Easier Question

A remarkable aspect of your mental life is that you are rarely stumped.True, you occasionally face a question such as 17 × 24 = ? to which noanswer comes immediately to mind, but these dumbfounded moments arerare. The normal state of your mind is that you have intuitive feelings andopinions about almost everything that comes your way. You like or dislikepeople long before you know much about them; you trust or distruststrangers without knowing why; you feel that an enterprise is bound tosucceed without analyzing it. Whether you state them or not, you often haveanswers to questions that you do not completely understand, relying onevidence that you can neither explain nor defend.

Substituting Questions

I propose a simple account of how we generate intuitive opinions oncomplex matters. If a satisfactory answer to a hard question isebr ques Dnot found quickly, System 1 will find a related question that is easier andwill answer it. I call the operation of answering one question in place ofanother substitution. I also adopt the following terms:

The target question is the assessment you intend to produce.The heuristic question is the simpler question that you answer instead.

The technical definition of heuristic is a simple procedure that helps findadequate, though often imperfect, answers to difficult questions. The wordcomes from the same root as eureka.

The idea of substitution came up early in my work with Amos, and it wasthe core of what became the heuristics and biases approach. We askedourselves how people manage to make judgments of probability withoutknowing precisely what probability is. We concluded that people mustsomehow simplify that impossible task, and we set out to find how they doit. Our answer was that when called upon to judge probability, peopleactually judge something else and believe they have judged probability.System 1 often makes this move when faced with difficult target questions,if the answer to a related and easier heuristic question comes readily tomind.

Substituting one question for another can be a good strategy for solvingdifficult problems, and George Pólya included substitution in his classic

(PDF) Thinking, fast and slow by daniel kahneman - PDFSLIDE.US (98)

How to Solve It: “If you can’t solve a problem, then there is an easierproblem you can solve: find it.” Pólya’s heuristics are strategic proceduresthat are deliberately implemented by System 2. But the heuristics that Idiscuss in this chapter are not chosen; they are a consequence of themental shotgun, the imprecise control we have over targeting ourresponses to questions.

Consider the questions listed in the left-hand column of table 1. Theseare difficult questions, and before you can produce a reasoned answer toany of them you must deal with other difficult issues. What is the meaningof happiness? What are the likely political developments in the next sixmonths? What are the standard sentences for other financial crimes? Howstrong is the competition that the candidate faces? What otherenvironmental or other causes should be considered? Dealing with thesequestions seriously is completely impractical. But you are not limited toperfectly reasoned answers to questions. There is a heuristic alternative tocareful reasoning, which sometimes works fairly well and sometimes leadsto serious errors.

Target Question Heuristic Question

How much would you contribute tosave an endangered species?

How much emotion do I feel whenI think of dying dolphins?

How happy are you with your lifethese days? What is my mood right now?

How popular is the president rightnow?

How popular will the president besix months from now?

How should financial advisers whoprey on the elderly be punished?

How much anger do I feel when Ithink of financial predators?

This woman is running for the primary.How far will she go in politics?

Does this woman look like apolitical winner?

Table 1

The mental shotgun makes it easy to generate quick answers to difficultquestions without imposing much hard work on your lazy System 2. The

(PDF) Thinking, fast and slow by daniel kahneman - PDFSLIDE.US (99)

right-hand counterpart of each of the left-hand questions is very likely to beevoked and very easily answered. Your feelings about dolphins andfinancial crooks, your current mood, your impressions of the political skill ofthe primary candidate, or the current standing of the president will readilycome to mind. The heuristic questions provide an off-the-shelf answer toeach of the difficult target questions.

Something is still missing from this story: the answers need to be fittedto the original questions. For example, my feelings about dying dolphinsmust be expressed in dollars. Another capability of System 1, intensitymatching, is available to solve that problem. Recall that both feelings andcontribution dollars are intensity scales. I can feel more or less stronglyabout dolphins and there is a contribution that matches the intensity of myfeelings. The dollar amount that will come to my mind is the matchingamount. Similar intensity matches are possible for all the questions. Forexample, the political skills of a candidate can range from pathetic toextraordinarily impressive, and the scale of political success can rangefrom the low of “She will be defeated in the primary” to a high of “She willsomeday be president of the United States.”

The automatic processes of the mental shotgun and intensity matchingoften make available one or more answers to easy questions that could bemapped onto the target question. On some occasions, substitution willoccur and a heuristic answer will be endorsed by System 2. Of course,System 2 has the opportunity to reject this intuitive answer, or to modify itby incorporating other information. However, a lazy System 2 often followsthe path of least effort and endorses a heuristic answer without muchscrutiny of whether it is truly appropriate. You will not be stumped, you willnot have to work very her р wheard, and you may not even notice that youdid not answer the question you were asked. Furthermore, you may notrealize that the target question was difficult, because an intuitive answer toit came readily to mind.

The 3-D Heuristic

Have a look at the picture of the three men and answer the question thatfollows.

(PDF) Thinking, fast and slow by daniel kahneman - PDFSLIDE.US (100)

Figure 9

As printed on the page, is the figure on the right larger than thefigure on the left?

The obvious answer comes quickly to mind: the figure on the right islarger. If you take a ruler to the two figures, however, you will discover thatin fact the figures are exactly the same size. Your impression of theirrelative size is dominated by a powerful illusion, which neatly illustrates theprocess of substitution.

The corridor in which the figures are seen is drawn in perspective andappears to go into the depth plane. Your perceptual system automaticallyinterprets the picture as a three-dimensional scene, not as an imageprinted on a flat paper surface. In the 3-D interpretation, the person on theright is both much farther away and much larger than the person on the left.For most of us, this impression of 3-D size is overwhelming. Only visual

(PDF) Thinking, fast and slow by daniel kahneman - PDFSLIDE.US (101)

artists and experienced photographers have developed the skill of seeingthe drawing as an object on the page. For the rest of us, substitutionoccurs: the dominant impression of 3-D size dictates the judgment of 2-Dsize. The illusion is due to a 3-D heuristic.

What happens here is a true illusion, not a misunderstanding of thequestion. You knew that the question was about the size of the figures inthe picture, as printed on the page. If you had been asked to estimate thesize of the figures, we know from experiments that your answer would havebeen in inches, not feet. You were not confused about the question, but youwere influenced by the answer to a question that you were not asked: “Howtall are the three people?”

The essential step in the heuristic—the substitution of three-dimensionalfor two-dimensional size—occurred automatically. The picture containscues that suggest a 3-D interpretation. These cues are irrelevant to thetask at hand—the judgment of size of the figure on the page—and youshould have ignored them, but you could not. The bias associated with theheuristic is that objects that appear to be more distant also appear to belarger on the page. As this example illustrates, a judgment that is based onsubstitution will inevitably be biased in predictable ways. In this case, ithappens so deep in the perceptual system that you simply cannot help it.

The Mood Heuristic for Happiness

A survey of German students is one of the best examples of substitution.The survey that the young participants completed included the followingtwo questions:

How happy are you these days?How many dates did you have last month?

< stрr to a p height="0%" width="0%">The experimenters were interestedin the correlation between the two answers. Would the students whoreported many dates say that they were happier than those with fewerdates? Surprisingly, no: the correlation between the answers was aboutzero. Evidently, dating was not what came first to the students’ minds whenthey were asked to assess their happiness. Another group of students sawthe same two questions, but in reverse order:

How many dates did you have last month?How happy are you these days?

The results this time were completely different. In this sequence, the

(PDF) Thinking, fast and slow by daniel kahneman - PDFSLIDE.US (102)

correlation between the number of dates and reported happiness wasabout as high as correlations between psychological measures can get.What happened?

The explanation is straightforward, and it is a good example ofsubstitution. Dating was apparently not the center of these students’ life (inthe first survey, happiness and dating were uncorrelated), but when theywere asked to think about their romantic life, they certainly had anemotional reaction. The students who had many dates were reminded of ahappy aspect of their life, while those who had none were reminded ofloneliness and rejection. The emotion aroused by the dating question wasstill on everyone’s mind when the query about general happiness came up.

The psychology of what happened is precisely analogous to thepsychology of the size illusion in figure 9. “Happiness these days” is not anatural or an easy assessment. A good answer requires a fair amount ofthinking. However, the students who had just been asked about their datingdid not need to think hard because they already had in their mind ananswer to a related question: how happy they were with their love life. Theysubstituted the question to which they had a readymade answer for thequestion they were asked.

Here again, as we did for the illusion, we can ask: Are the studentsconfused? Do they really think that the two questions—the one they wereasked and the one they answer—are synonymous? Of course not. Thestudents do not temporarily lose their ability to distinguish romantic lifefrom life as a whole. If asked about the two concepts, they would say theyare different. But they were not asked whether the concepts are different.They were asked how happy they were, and System 1 has a ready answer.

Dating is not unique. The same pattern is found if a question about thestudents’ relations with their parents or about their finances immediatelyprecedes the question about general happiness. In both cases,satisfaction in the particular domain dominates happiness reports. Anyemotionally significant question that alters a person’s mood will have thesame effect. WYSIATI. The present state of mind looms very large whenpeople evaluate their happiness.

The Affect Heuristic

The dominance of conclusions over arguments is most pronounced whereemotions are involved. The psychologist Paul Slovic has proposed anaffect heuristic in which people let their likes and dislikes determine theirbeliefs about the world. Your political preference determines thearguments that you find compelling. If you like the current health policy, you

(PDF) Thinking, fast and slow by daniel kahneman - PDFSLIDE.US (103)

believe its benefits are substantial and its costs more manageable thanthe costs of alternatives. If you are a hawk in your attitude toward othernations, you probabltheр"0%y think they are relatively weak and likely tosubmit to your country’s will. If you are a dove, you probably think they arestrong and will not be easily coerced. Your emotional attitude to suchthings as irradiated food, red meat, nuclear power, tattoos, or motorcyclesdrives your beliefs about their benefits and their risks. If you dislike any ofthese things, you probably believe that its risks are high and its benefitsnegligible.

The primacy of conclusions does not mean that your mind is completelyclosed and that your opinions are wholly immune to information andsensible reasoning. Your beliefs, and even your emotional attitude, maychange (at least a little) when you learn that the risk of an activity youdisliked is smaller than you thought. However, the information about lowerrisks will also change your view of the benefits (for the better) even ifnothing was said about benefits in the information you received.

We see here a new side of the “personality” of System 2. Until now Ihave mostly described it as a more or less acquiescent monitor, whichallows considerable leeway to System 1. I have also presented System 2as active in deliberate memory search, complex computations,comparisons, planning, and choice. In the bat-and-ball problem and inmany other examples of the interplay between the two systems, itappeared that System 2 is ultimately in charge, with the ability to resist thesuggestions of System 1, slow things down, and impose logical analysis.Self-criticism is one of the functions of System 2. In the context of attitudes,however, System 2 is more of an apologist for the emotions of System 1than a critic of those emotions—an endorser rather than an enforcer. Itssearch for information and arguments is mostly constrained to informationthat is consistent with existing beliefs, not with an intention to examinethem. An active, coherence-seeking System 1 suggests solutions to anundemanding System 2.

Speaking of Substitution and Heuristics

“Do we still remember the question we are trying to answer? Orhave we substituted an easier one?”

“The question we face is whether this candidate can succeed.The question we seem to answer is whether she interviews well.Let’s not substitute.”

(PDF) Thinking, fast and slow by daniel kahneman - PDFSLIDE.US (104)

“He likes the project, so he thinks its costs are low and itsbenefits are high. Nice example of the affect heuristic.”

“We are using last year’s performance as a heuristic to predictthe value of the firm several years from now. Is this heuristic goodenough? What other information do we need?”

The table below contains a list of features and activities that have beenattributed to System 1. Each of the active sentences replaces a statement,technically more accurate but harder to understand, to the effect that amental event occurs automatically and fast. My hope is that the list of traitswill help you develop an intuitive sense of the “personality” of the fictitiousSystem 1. As happens with other characters you know, you will havehunches about what System 1 would do under different circumstances, andmost of your hunches will be correct.

Characteristics of System 1

generates impressions, feelings, and inclinations; when endorsed bySystem 2 these become beliefs, attitudes, and intentionsoperates automatically and quickly, with little or no effort, and nosense of voluntary controlcan be programmed by System 2 to mobilize attention when aparticular pattern is detected (search)executes skilled responses and generates skilled intuitions, afteradequate trainingcreates a coherent pattern of activated ideas in associative memorylinks a sense of cognitive ease to illusions of truth, pleasant feelings,and reduced vigilancedistinguishes the surprising from the normalinfers and invents causes and intentionsneglects ambiguity and suppresses doubtis biased to believe and confirmexaggerates emotional consistency (halo effect)focuses on existing evidence and ignores absent evidence

(PDF) Thinking, fast and slow by daniel kahneman - PDFSLIDE.US (105)

(WYSIATI)

generates a limited set of basic assessmentsrepresents sets by norms and prototypes, does not integrate

matches intensities across scales (e.g., size to loudness)computes more than intended (mental shotgun)sometimes substitutes an easier question for a difficult one(heuristics)is more sensitive to changes than to states (prospect theory)*

overweights low probabilities*

shows diminishing sensitivity to quantity (psychophysics)*

responds more strongly to losses than to gains (loss aversion)*

frames decision problems narrowly, in isolation from one another*

(PDF) Thinking, fast and slow by daniel kahneman - PDFSLIDE.US (106)

Part 2

(PDF) Thinking, fast and slow by daniel kahneman - PDFSLIDE.US (107)

Heuristics and Biases

(PDF) Thinking, fast and slow by daniel kahneman - PDFSLIDE.US (108)

The Law of Small Numbers

A study of the incidence of kidney cancer in the 3,141 counties of theUnited a>< HЉStates reveals a remarkable pattern. The counties in whichthe incidence of kidney cancer is lowest are mostly rural, sparselypopulated, and located in traditionally Republican states in the Midwest,the South, and the West. What do you make of this?

Your mind has been very active in the last few seconds, and it wasmainly a System 2 operation. You deliberately searched memory andformulated hypotheses. Some effort was involved; your pupils dilated, andyour heart rate increased measurably. But System 1 was not idle: theoperation of System 2 depended on the facts and suggestions retrievedfrom associative memory. You probably rejected the idea that Republicanpolitics provide protection against kidney cancer. Very likely, you ended upfocusing on the fact that the counties with low incidence of cancer aremostly rural. The witty statisticians Howard Wainer and Harris Zwerling,from whom I learned this example, commented, “It is both easy andtempting to infer that their low cancer rates are directly due to the cleanliving of the rural lifestyle—no air pollution, no water pollution, access tofresh food without additives.” This makes perfect sense.

Now consider the counties in which the incidence of kidney cancer ishighest. These ailing counties tend to be mostly rural, sparsely populated,and located in traditionally Republican states in the Midwest, the South,and the West. Tongue-in-cheek, Wainer and Zwerling comment: “It is easyto infer that their high cancer rates might be directly due to the poverty ofthe rural lifestyle—no access to good medical care, a high-fat diet, and toomuch alcohol, too much tobacco.” Something is wrong, of course. The rurallifestyle cannot explain both very high and very low incidence of kidneycancer.

The key factor is not that the counties were rural or predominantlyRepublican. It is that rural counties have small populations. And the mainlesson to be learned is not about epidemiology, it is about the difficultrelationship between our mind and statistics. System 1 is highly adept inone form of thinking—it automatically and effortlessly identifies causalconnections between events, sometimes even when the connection isspurious. When told about the high-incidence counties, you immediatelyassumed that these counties are different from other counties for a reason,that there must be a cause that explains this difference. As we shall see,however, System 1 is inept when faced with “merely statistical” facts, whichchange the probability of outcomes but do not cause them to happen.

A random event, by definition, does not lend itself to explanation, but

(PDF) Thinking, fast and slow by daniel kahneman - PDFSLIDE.US (109)

collections of random events do behave in a highly regular fashion.Imagine a large urn filled with marbles. Half the marbles are red, half arewhite. Next, imagine a very patient person (or a robot) who blindly draws 4marbles from the urn, records the number of red balls in the sample, throwsthe balls back into the urn, and then does it all again, many times. If yousummarize the results, you will find that the outcome “2 red, 2 white” occurs(almost exactly) 6 times as often as the outcome “4 red” or “4 white.” Thisrelationship is a mathematical fact. You can predict the outcome ofrepeated sampling from an urn just as confidently as you can predict whatwill happen if you hit an egg with a hammer. You cannot predict every detailof how the shell will shatter, but you can be sure of the general idea. Thereis a difference: the satisfying sense of causation that you experience whenthinking of a hammer hitting an egg is altogether absent when you thinkabout sampling.

A related statistical fact is relevant to the cancer example. From thesame urn, two very patient marble counters thatрy dake turns. Jack draws4 marbles on each trial, Jill draws 7. They both record each time theyobserve a homogeneous sample—all white or all red. If they go on longenough, Jack will observe such extreme outcomes more often than Jill—bya factor of 8 (the expected percentages are 12.5% and 1.56%). Again, nohammer, no causation, but a mathematical fact: samples of 4 marblesyield extreme results more often than samples of 7 marbles do.

Now imagine the population of the United States as marbles in a gianturn. Some marbles are marked KC, for kidney cancer. You draw samplesof marbles and populate each county in turn. Rural samples are smallerthan other samples. Just as in the game of Jack and Jill, extremeoutcomes (very high and/or very low cancer rates) are most likely to befound in sparsely populated counties. This is all there is to the story.

We started from a fact that calls for a cause: the incidence of kidneycancer varies widely across counties and the differences are systematic.The explanation I offered is statistical: extreme outcomes (both high andlow) are more likely to be found in small than in large samples. Thisexplanation is not causal. The small population of a county neither causesnor prevents cancer; it merely allows the incidence of cancer to be muchhigher (or much lower) than it is in the larger population. The deeper truth isthat there is nothing to explain. The incidence of cancer is not truly lower orhigher than normal in a county with a small population, it just appears to beso in a particular year because of an accident of sampling. If we repeat theanalysis next year, we will observe the same general pattern of extremeresults in the small samples, but the counties where cancer was commonlast year will not necessarily have a high incidence this year. If this is thecase, the differences between dense and rural counties do not really count

(PDF) Thinking, fast and slow by daniel kahneman - PDFSLIDE.US (110)

as facts: they are what scientists call artifacts, observations that areproduced entirely by some aspect of the method of research—in this case,by differences in sample size.

The story I have told may have surprised you, but it was not a revelation.You have long known that the results of large samples deserve more trustthan smaller samples, and even people who are innocent of statisticalknowledge have heard about this law of large numbers. But “knowing” isnot a yes-no affair and you may find that the following statements apply toyou:

The feature “sparsely populated” did not immediately stand out asrelevant when you read the epidemiological story.You were at least mildly surprised by the size of the differencebetween samples of 4 and samples of 7.Even now, you must exert some mental effort to see that the followingtwo statements mean exactly the same thing:

Large samples are more precise than small samples.Small samples yield extreme results more often than largesamples do.

The first statement has a clear ring of truth, but until the second versionmakes intuitive sense, you have not truly understood the first.

The bottom line: yes, you did know that the results of large samples aremore precise, but you may now realize that you did not know it very well.You are not alone. The first study that Amos and I did together showed thateven sophisticated researchers have poor intuitions and a wobblyunderstanding of sampling effects.

The Law of Small Numbers

My collaboration with Amos in the early 1970s began with a discussion ofthe claim that people who have had no training in statistics are good“intuitive statisticians.” He told my seminar and me of researchers at theUniversity of Michigan who were generally optimistic about intuitivestatistics. I had strong feelings about that claim, which I took personally: Ihad recently discovered that I was not a good intuitive statistician, and I didnot believe that I was worse than others.

For a research psychologist, sampling variation is not a curiosity; it is anuisance and a costly obstacle, which turns the undertaking of every

(PDF) Thinking, fast and slow by daniel kahneman - PDFSLIDE.US (111)

research project into a gamble. Suppose that you wish to confirm thehypothesis that the vocabulary of the average six-year-old girl is larger thanthe vocabulary of an average boy of the same age. The hypothesis is truein the population; the average vocabulary of girls is indeed larger. Girls andboys vary a great deal, however, and by the luck of the draw you couldselect a sample in which the difference is inconclusive, or even one inwhich boys actually score higher. If you are the researcher, this outcome iscostly to you because you have wasted time and effort, and failed toconfirm a hypothesis that was in fact true. Using a sufficiently large sampleis the only way to reduce the risk. Researchers who pick too small asample leave themselves at the mercy of sampling luck.

The risk of error can be estimated for any given sample size by a fairlysimple procedure. Traditionally, however, psychologists do not usecalculations to decide on a sample size. They use their judgment, which iscommonly flawed. An article I had read shortly before the debate withAmos demonstrated the mistake that researchers made (they still do) by adramatic observation. The author pointed out that psychologists commonlychose samples so small that they exposed themselves to a 50% risk offailing to confirm their true hypotheses! No researcher in his right mindwould accept such a risk. A plausible explanation was that psychologists’decisions about sample size reflected prevalent intuitive misconceptionsof the extent of sampling variation.

The article shocked me, because it explained some troubles I had had inmy own research. Like most research psychologists, I had routinely chosensamples that were too small and had often obtained results that made nosense. Now I knew why: the odd results were actually artifacts of myresearch method. My mistake was particularly embarrassing because Itaught statistics and knew how to compute the sample size that wouldreduce the risk of failure to an acceptable level. But I had never chosen asample size by computation. Like my colleagues, I had trusted traditionand my intuition in planning my experiments and had never thoughtseriously about the issue. When Amos visited the seminar, I had alreadyreached the conclusion that my intuitions were deficient, and in the courseof the seminar we quickly agreed that the Michigan optimists were wrong.

Amos and I set out to examine whether I was the only fool or a memberof a majority of fools, by testing whether researchers selected formathematical expertise would make similar mistakes. We developed aquestionnaire that described realistic research situations, includingreplications of successful experiments. It asked the researchers to choosesample sizes, to assess the risks of failure to which their decisionsexposed them, and to provide advice to hypothetical graduate studentsplanning their research. Amos collected the responses of a group of

(PDF) Thinking, fast and slow by daniel kahneman - PDFSLIDE.US (112)

sophisticated participants (including authors of two statistical textbooks) ata meetatiрp>

Amos and I called our first joint article “Belief in the Law of SmallNumbers.” We explained, tongue-in-cheek, that “intuitions about randomsampling appear to satisfy the law of small numbers, which asserts that thelaw of large numbers applies to small numbers as well.” We also includeda strongly worded recommendation that researchers regard their“statistical intuitions with proper suspicion and replace impressionformation by computation whenever possible.”

A Bias of Confidence Over Doubt

In a telephone poll of 300 seniors, 60% support the president.

If you had to summarize the message of this sentence in exactly threewords, what would they be? Almost certainly you would choose “elderlysupport president.” These words provide the gist of the story. The omitteddetails of the poll, that it was done on the phone with a sample of 300, areof no interest in themselves; they provide background information thatattracts little attention. Your summary would be the same if the sample sizehad been different. Of course, a completely absurd number would drawyour attention (“a telephone poll of 6 [or 60 million] elderly voters…”).Unless you are a professional, however, you may not react very differentlyto a sample of 150 and to a sample of 3,000. That is the meaning of thestatement that “people are not adequately sensitive to sample size.”

The message about the poll contains information of two kinds: the storyand the source of the story. Naturally, you focus on the story rather than onthe reliability of the results. When the reliability is obviously low, however,the message will be discredited. If you are told that “a partisan group hasconducted a flawed and biased poll to show that the elderly support thepresident…” you will of course reject the findings of the poll, and they willnot become part of what you believe. Instead, the partisan poll and its falseresults will become a new story about political lies. You can choose todisbelieve a message in such clear-cut cases. But do you discriminatesufficiently between “I read in The New York Times…” and “I heard at thewatercooler…”? Can your System 1 distinguish degrees of belief? Theprinciple of WY SIATI suggests that it cannot.

As I described earlier, System 1 is not prone to doubt. It suppressesambiguity and spontaneously constructs stories that are as coherent aspossible. Unless the message is immediately negated, the associations

(PDF) Thinking, fast and slow by daniel kahneman - PDFSLIDE.US (113)

that it evokes will spread as if the message were true. System 2 is capableof doubt, because it can maintain incompatible possibilities at the sametime. However, sustaining doubt is harder work than sliding into certainty.The law of small numbers is a manifestation of a general bias that favorscertainty over doubt, which will turn up in many guises in following chapters.

The strong bias toward believing that small samples closely resemblethe population from which they are drawn is also part of a larger story: weare prone to exaggerate the consistency and coherence of what we see.The exaggerated faith of researchers in what can be learned from a fewobservations is closely related to the halo effect thрhe , the sense we oftenget that we know and understand a person about whom we actually knowvery little. System 1 runs ahead of the facts in constructing a rich image onthe basis of scraps of evidence. A machine for jumping to conclusions willact as if it believed in the law of small numbers. More generally, it willproduce a representation of reality that makes too much sense.

Cause and Chance

The associative machinery seeks causes. The difficulty we have withstatistical regularities is that they call for a different approach. Instead offocusing on how the event at hand came to be, the statistical view relates itto what could have happened instead. Nothing in particular caused it to bewhat it is—chance selected it from among its alternatives.

Our predilection for causal thinking exposes us to serious mistakes inevaluating the randomness of truly random events. For an example, takethe sex of six babies born in sequence at a hospital. The sequence of boysand girls is obviously random; the events are independent of each other,and the number of boys and girls who were born in the hospital in the lastfew hours has no effect whatsoever on the sex of the next baby. Nowconsider three possible sequences:

BBBGGGGGGGGGBGBBGB

Are the sequences equally likely? The intuitive answer—“of course not!”—is false. Because the events are independent and because the outcomesB and G are (approximately) equally likely, then any possible sequence ofsix births is as likely as any other. Even now that you know this conclusionis true, it remains counterintuitive, because only the third sequenceappears random. As expected, BGBBGB is judged much more likely than

(PDF) Thinking, fast and slow by daniel kahneman - PDFSLIDE.US (114)

the other two sequences. We are pattern seekers, believers in a coherentworld, in which regularities (such as a sequence of six girls) appear not byaccident but as a result of mechanical causality or of someone’s intention.We do not expect to see regularity produced by a random process, andwhen we detect what appears to be a rule, we quickly reject the idea thatthe process is truly random. Random processes produce many sequencesthat convince people that the process is not random after all. You can seewhy assuming causality could have had evolutionary advantages. It is partof the general vigilance that we have inherited from ancestors. We areautomatically on the lookout for the possibility that the environment haschanged. Lions may appear on the plain at random times, but it would besafer to notice and respond to an apparent increase in the rate ofappearance of prides of lions, even if it is actually due to the fluctuations ofa random process.

The widespread misunderstanding of randomness sometimes hassignificant consequences. In our article on representativeness, Amos and Icited the statistician William Feller, who illustrated the ease with whichpeople see patterns where none exists. During the intensive rocketbombing of London in World War II, it was generally believed that thebombing could not be random because a map of the hits revealedconspicuous gaps. Some suspected that German spies were located inthe unharmed areas. A careful statistical analysis revealed that thedistribution of hits was typical of a random process—and typical as well inevoking a strong impression that it was not random. “To the untrained eye,”Feller remarks, “randomness appears as regularity or tendency to cluster.”

I soon had an occasion to apply what I had learned frpeaрrainom Feller.The Yom Kippur War broke out in 1973, and my only significantcontribution to the war effort was to advise high officers in the Israeli AirForce to stop an investigation. The air war initially went quite badly forIsrael, because of the unexpectedly good performance of Egyptian ground-to-air missiles. Losses were high, and they appeared to be unevenlydistributed. I was told of two squadrons flying from the same base, one ofwhich had lost four planes while the other had lost none. An inquiry wasinitiated in the hope of learning what it was that the unfortunate squadronwas doing wrong. There was no prior reason to believe that one of thesquadrons was more effective than the other, and no operationaldifferences were found, but of course the lives of the pilots differed in manyrandom ways, including, as I recall, how often they went home betweenmissions and something about the conduct of debriefings. My advice wasthat the command should accept that the different outcomes were due toblind luck, and that the interviewing of the pilots should stop. I reasonedthat luck was the most likely answer, that a random search for a

(PDF) Thinking, fast and slow by daniel kahneman - PDFSLIDE.US (115)

nonobvious cause was hopeless, and that in the meantime the pilots in thesquadron that had sustained losses did not need the extra burden of beingmade to feel that they and their dead friends were at fault.

Some years later, Amos and his students Tom Gilovich and RobertVallone caused a stir with their study of misperceptions of randomness inbasketball. The “fact” that players occasionally acquire a hot hand isgenerally accepted by players, coaches, and fans. The inference isirresistible: a player sinks three or four baskets in a row and you cannothelp forming the causal judgment that this player is now hot, with atemporarily increased propensity to score. Players on both teams adapt tothis judgment—teammates are more likely to pass to the hot scorer andthe defense is more likely to doubleteam. Analysis of thousands ofsequences of shots led to a disappointing conclusion: there is no suchthing as a hot hand in professional basketball, either in shooting from thefield or scoring from the foul line. Of course, some players are moreaccurate than others, but the sequence of successes and missed shotssatisfies all tests of randomness. The hot hand is entirely in the eye of thebeholders, who are consistently too quick to perceive order and causalityin randomness. The hot hand is a massive and widespread cognitiveillusion.

The public reaction to this research is part of the story. The finding waspicked up by the press because of its surprising conclusion, and thegeneral response was disbelief. When the celebrated coach of the BostonCeltics, Red Auerbach, heard of Gilovich and his study, he responded,“Who is this guy? So he makes a study. I couldn’t care less.” The tendencyto see patterns in randomness is overwhelming—certainly moreimpressive than a guy making a study.

The illusion of pattern affects our lives in many ways off the basketballcourt. How many good years should you wait before concluding that aninvestment adviser is unusually skilled? How many successful acquisitionsshould be needed for a board of directors to believe that the CEO hasextraordinary flair for such deals? The simple answer to these questions isthat if you follow your intuition, you will more often than not err bymisclassifying a random event as systematic. We are far too willing toreject the belief that much of what we see in life is random.

I began this chapter with the example of cancer incidence across theUnited States. The example appears in a book intended for statisticsteachers, but I learned about it from an amusing article by the twostatisticians I quoted earlier, Howard Wainer and Harris Zwerling. Theiressay focused on a large iiveрothersnvestment, some $1.7 billion, whichthe Gates Foundation made to follow up intriguing findings on the

(PDF) Thinking, fast and slow by daniel kahneman - PDFSLIDE.US (116)

characteristics of the most successful schools. Many researchers havesought the secret of successful education by identifying the mostsuccessful schools in the hope of discovering what distinguishes themfrom others. One of the conclusions of this research is that the mostsuccessful schools, on average, are small. In a survey of 1,662 schools inPennsylvania, for instance, 6 of the top 50 were small, which is anoverrepresentation by a factor of 4. These data encouraged the GatesFoundation to make a substantial investment in the creation of smallschools, sometimes by splitting large schools into smaller units. At leasthalf a dozen other prominent institutions, such as the AnnenbergFoundation and the Pew Charitable Trust, joined the effort, as did the U.S.Department of Education’s Smaller Learning Communities Program.

This probably makes intuitive sense to you. It is easy to construct acausal story that explains how small schools are able to provide superioreducation and thus produce high-achieving scholars by giving them morepersonal attention and encouragement than they could get in largerschools. Unfortunately, the causal analysis is pointless because the factsare wrong. If the statisticians who reported to the Gates Foundation hadasked about the characteristics of the worst schools, they would havefound that bad schools also tend to be smaller than average. The truth isthat small schools are not better on average; they are simply morevariable. If anything, say Wainer and Zwerling, large schools tend toproduce better results, especially in higher grades where a variety ofcurricular options is valuable.

Thanks to recent advances in cognitive psychology, we can now seeclearly what Amos and I could only glimpse: the law of small numbers ispart of two larger stories about the workings of the mind.

The exaggerated faith in small samples is only one example of amore general illusion—we pay more attention to the content ofmessages than to information about their reliability, and as a resultend up with a view of the world around us that is simpler and morecoherent than the data justify. Jumping to conclusions is a safer sportin the world of our imagination than it is in reality.Statistics produce many observations that appear to beg for causalexplanations but do not lend themselves to such explanations. Manyfacts of the world are due to chance, including accidents of sampling.Causal explanations of chance events are inevitably wrong.

(PDF) Thinking, fast and slow by daniel kahneman - PDFSLIDE.US (117)

Speaking of the Law of Small Numbers

“Yes, the studio has had three successful films since the newCEO took over. But it is too early to declare he has a hot hand.”

“I won’t believe that the new trader is a genius before consulting astatistician who could estimate the likelihood of his streak beinga chance event.”

“The sample of observations is too small to make any inferences.Let’s not follow the law of small numbers.”

“I plan to keep the results of the experiment secret until we have asufficiently large sample. Otherwisortрxpere we will face pressureto reach a conclusion prematurely.”

(PDF) Thinking, fast and slow by daniel kahneman - PDFSLIDE.US (118)

Anchors

Amos and I once rigged a wheel of fortune. It was marked from 0 to 100,but we had it built so that it would stop only at 10 or 65. We recruitedstudents of the University of Oregon as participants in our experiment. Oneof us would stand in front of a small group, spin the wheel, and ask them towrite down the number on which the wheel stopped, which of course waseither 10 or 65. We then asked them two questions:

Is the percentage of African nations among UN members largeror smaller than the number you just wrote?

What is your best guess of the percentage of African nations inthe UN?

The spin of a wheel of fortune—even one that is not rigged—cannotpossibly yield useful information about anything, and the participants in ourexperiment should simply have ignored it. But they did not ignore it. Theaverage estimates of those who saw 10 and 65 were 25% and 45%,respectively.

The phenomenon we were studying is so common and so important inthe everyday world that you should know its name: it is an anchoring effect.It occurs when people consider a particular value for an unknown quantitybefore estimating that quantity. What happens is one of the most reliableand robust results of experimental psychology: the estimates stay close tothe number that people considered—hence the image of an anchor. If youare asked whether Gandhi was more than 114 years old when he died youwill end up with a much higher estimate of his age at death than you wouldif the anchoring question referred to death at 35. If you consider how muchyou should pay for a house, you will be influenced by the asking price. Thesame house will appear more valuable if its listing price is high than if it islow, even if you are determined to resist the influence of this number; andso on—the list of anchoring effects is endless. Any number that you areasked to consider as a possible solution to an estimation problem willinduce an anchoring effect.

We were not the first to observe the effects of anchors, but ourexperiment was the first demonstration of its absurdity: people’s judgmentswere influenced by an obviously uninformative number. There was no wayto describe the anchoring effect of a wheel of fortune as reasonable. Amosand I published the experiment in our Science paper, and it is one of the

(PDF) Thinking, fast and slow by daniel kahneman - PDFSLIDE.US (119)

best known of the findings we reported there.There was only one trouble: Amos and I did not fully agree on the

psychology of the anchoring effect. He supported one interpretation, I likedanother, and we never found a way to settle the argument. The problemwas finally solved decades later by the efforts of numerous investigators. Itis now clear that Amos and I were both right. Two different mechanismsproduce anchoring effects—one for each system. There is a form ofanchoring that occurs in a deliberate process of adjustment, an operationof System 2. And there is anchoring that occurs by a priming effect, anautomatic manifestation of System 1.

Anchoring as Adjustment

Amos liked the idea of an adjust-and-anchor heuristic as a strategy forestimating uncertain quantities: start from an anchoring number, assesswhether it is too high or too low, and gradually adjust your estimate bymentally “moving” from the anchor. The adjustment typically endsprematurely, because people stop when they are no longer certain thatthey should move farther. Decades after our disagreement, and years afterAmos’s death, convincing evidence of such a process was offeredindependently by two psychologists who had worked closely with Amosearly in their careers: Eldar Shafir and Tom Gilovich together with their ownstudents—Amos’s intellectual grandchildren!

To get the idea, take a sheet of paper and draw a 2½-inch line going up,starting at the bottom of the page—without a ruler. Now take another sheet,and start at the top and draw a line going down until it is 2½ inches fromthe bottom. Compare the lines. There is a good chance that your firstestimate of 2½ inches was shorter than the second. The reason is that youdo not know exactly what such a line looks like; there is a range ofuncertainty. You stop near the bottom of the region of uncertainty when youstart from the bottom of the page and near the top of the region when youstart from the top. Robyn Le Boeuf and Shafir found many examples of thatmechanism in daily experience. Insufficient adjustment neatly explains whyyou are likely to drive too fast when you come off the highway onto citystreets—especially if you are talking with someone as you drive.Insufficient adjustment is also a source of tension between exasperatedparents and teenagers who enjoy loud music in their room. Le Boeuf andShafir note that a “well-intentioned child who turns down exceptionally loudmusic to meet a parent’s demand that it be played at a ‘reasonable’volume may fail to adjust sufficiently from a high anchor, and may feel thatgenuine attempts at compromise are being overlooked.” The driver and

(PDF) Thinking, fast and slow by daniel kahneman - PDFSLIDE.US (120)

the child both deliberately adjust down, and both fail to adjust enough.Now consider these questions:

When did George Washington become president?What is the boiling temperature of water at the top of MountEverest?

The first thing that happens when you consider each of these questions isthat an anchor comes to your mind, and you know both that it is wrong andthe direction of the correct answer. You know immediately that GeorgeWashington became president after 1776, and you also know that theboiling temperature of water at the top of Mount Everest is lower than100°C. You have to adjust in the appropriate direction by findingarguments to move away from the anchor. As in the case of the lines, youare likely to stop when you are no longer sure you should go farther—at thenear edge of the region of uncertainty.

Nick Epley and Tom Gilovich found evidence that adjustment is adeliberate attempt to find reasons to move away from the anchor: peoplewho are instructed to shake their head when they hear the anchor, as ifthey rejected it, move farther from the anchor, and people who nod theirhead show enhanced anchoring. Epley and Gilovich also confirmed thatadjustment is an effortful operation. People adjust less (stay closer to theanchor) when their mental resources are depleted, either because theirmemory is loaded with dighdth=igits or because they are slightly drunk.Insufficient adjustment is a failure of a weak or lazy System 2.

So we now know that Amos was right for at least some cases ofanchoring, which involve a deliberate System 2 adjustment in a specifieddirection from an anchor.

Anchoring as Priming Effect

When Amos and I debated anchoring, I agreed that adjustment sometimesoccurs, but I was uneasy. Adjustment is a deliberate and consciousactivity, but in most cases of anchoring there is no correspondingsubjective experience. Consider these two questions:

Was Gandhi more or less than 144 years old when he died?How old was Gandhi when he died?

Did you produce your estimate by adjusting down from 144? Probably not,

(PDF) Thinking, fast and slow by daniel kahneman - PDFSLIDE.US (121)

but the absurdly high number still affected your estimate. My hunch was thatanchoring is a case of suggestion. This is the word we use when someonecauses us to see, hear, or feel something by merely bringing it to mind. Forexample, the question “Do you now feel a slight numbness in your left leg?”always prompts quite a few people to report that their left leg does indeedfeel a little strange.

Amos was more conservative than I was about hunches, and he correctlypointed out that appealing to suggestion did not help us understandanchoring, because we did not know how to explain suggestion. I had toagree that he was right, but I never became enthusiastic about the idea ofinsufficient adjustment as the sole cause of anchoring effects. Weconducted many inconclusive experiments in an effort to understandanchoring, but we failed and eventually gave up the idea of writing moreabout it.

The puzzle that defeated us is now solved, because the concept ofsuggestion is no longer obscure: suggestion is a priming effect, whichselectively evokes compatible evidence. You did not believe for a momentthat Gandhi lived for 144 years, but your associative machinery surelygenerated an impression of a very ancient person. System 1 understandssentences by trying to make them true, and the selective activation ofcompatible thoughts produces a family of systematic errors that make usgullible and prone to believe too strongly whatever we believe. We can nowsee why Amos and I did not realize that there were two types of anchoring:the research techniques and theoretical ideas we needed did not yet exist.They were developed, much later, by other people. A process thatresembles suggestion is indeed at work in many situations: System 1 triesits best to construct a world in which the anchor is the true number. This isone of the manifestations of associative coherence that I described in thefirst part of the book.

The German psychologists Thomas Mussweiler and Fritz Strack offeredthe most compelling demonstrations of the role of associative coherencein anchoring. In one experiment, they asked an anchoring question abouttemperature: “Is the annual mean temperature in Germany higher or lowerthan 20°C (68°F)?” or “Is the annual mean temperature in Germany higheror lower than 5°C (40°F)?”

All participants were then briefly shown words that they were asked toidentify. The researchers found that 68°F made it easier to recognizesummer words (like sun and beach), and 40°F facilitated winter words(like frost and ski). The selective activation of compatible memoriesexplains anchoring: the high and the low numbers activate different sets ofideas in memory. The estimates of annual temperature draw on these

(PDF) Thinking, fast and slow by daniel kahneman - PDFSLIDE.US (122)

biased samples of ideas and are therefore biased as well. In anotherelegant study in the same vein, participants were asked about the averageprice of German cars. A high anchor selectively primed the names of luxurybrands (Mercedes, Audi), whereas the low anchor primed brandsassociated with mass-market cars (Volkswagen). We saw earlier that anyprime will tend to evoke information that is compatible with it. Suggestionand anchoring are both explained by the same automatic operation ofSystem 1. Although I did not know how to prove it at the time, my hunchabout the link between anchoring and suggestion turned out to be correct.

The Anchoring Index

Many psychological phenomena can be demonstrated experimentally, butfew can actually be measured. The effect of anchors is an exception.Anchoring can be measured, and it is an impressively large effect. Somevisitors at the San Francisco Exploratorium were asked the following twoquestions:

Is the height of the tallest redwood more or less than 1,200 feet?What is your best guess about the height of the tallest redwood?

The “high anchor” in this experiment was 1,200 feet. For other participants,the first question referred to a “low anchor” of 180 feet. The differencebetween the two anchors was 1,020 feet.

As expected, the two groups produced very different mean estimates:844 and 282 feet. The difference between them was 562 feet. Theanchoring index is simply the ratio of the two differences (562/1,020)expressed as a percentage: 55%. The anchoring measure would be 100%for people who slavishly adopt the anchor as an estimate, and zero forpeople who are able to ignore the anchor altogether. The value of 55% thatwas observed in this example is typical. Similar values have beenobserved in numerous other problems.

The anchoring effect is not a laboratory curiosity; it can be just as strongin the real world. In an experiment conducted some years ago, real-estateagents were given an opportunity to assess the value of a house that wasactually on the market. They visited the house and studied acomprehensive booklet of information that included an asking price. Halfthe agents saw an asking price that was substantially higher than the listedprice of the house; the other half saw an asking price that was substantiallylower. Each agent gave her opinion about a reasonable buying price forthe house and the lowest price at which she would agree to sell the houseif she owned it. The agents were then asked about the factors that had

(PDF) Thinking, fast and slow by daniel kahneman - PDFSLIDE.US (123)

affected their judgment. Remarkably, the asking price was not one of thesefactors; the agents took pride in their ability to ignore it. They insisted thatthe listing price had no effect on their responses, but they were wrong: theanchoring effect was 41%. Indeed, the professionals were almost assusceptible to anchoring effects as business school students with no real-estate experience, whose anchoring index was 48%. The only differencebetween the two groups was that the students conceded that they wereinfluenced by the anchor, while the professionals denied that influence.

Powerful anchoring effects are found in decisions that people makeabout money, such as when they choose how much to contribute al.lsdenied to a cause. To demonstrate this effect, we told participants in theExploratorium study about the environmental damage caused by oiltankers in the Pacific Ocean and asked about their willingness to make anannual contribution “to save 50,000 offshore Pacific Coast seabirds fromsmall offshore oil spills, until ways are found to prevent spills or requiretanker owners to pay for the operation.” This question requires intensitymatching: the respondents are asked, in effect, to find the dollar amount ofa contribution that matches the intensity of their feelings about the plight ofthe seabirds. Some of the visitors were first asked an anchoring question,such as, “Would you be willing to pay $5…,” before the point-blankquestion of how much they would contribute.

When no anchor was mentioned, the visitors at the Exploratorium—generally an environmentally sensitive crowd—said they were willing to pay$64, on average. When the anchoring amount was only $5, contributionsaveraged $20. When the anchor was a rather extravagant $400, thewillingness to pay rose to an average of $143.

The difference between the high-anchor and low-anchor groups was$123. The anchoring effect was above 30%, indicating that increasing theinitial request by $100 brought a return of $30 in average willingness topay.

Similar or even larger anchoring effects have been obtained innumerous studies of estimates and of willingness to pay. For example,French residents of the heavily polluted Marseilles region were asked whatincrease in living costs they would accept if they could live in a lesspolluted region. The anchoring effect was over 50% in that study.Anchoring effects are easily observed in online trading, where the sameitem is often offered at different “buy now” prices. The “estimate” in fine-artauctions is also an anchor that influences the first bid.

There are situations in which anchoring appears reasonable. After all, itis not surprising that people who are asked difficult questions clutch atstraws, and the anchor is a plausible straw. If you know next to nothing

(PDF) Thinking, fast and slow by daniel kahneman - PDFSLIDE.US (124)

about the trees of California and are asked whether a redwood can betaller than 1,200 feet, you might infer that this number is not too far from thetruth. Somebody who knows the true height thought up that question, so theanchor may be a valuable hint. However, a key finding of anchoringresearch is that anchors that are obviously random can be just as effectiveas potentially informative anchors. When we used a wheel of fortune toanchor estimates of the proportion of African nations in the UN, theanchoring index was 44%, well within the range of effects observed withanchors that could plausibly be taken as hints. Anchoring effects of similarsize have been observed in experiments in which the last few digits of therespondent’s Social Security number was used as the anchor (e.g., forestimating the number of physicians in their city). The conclusion is clear:anchors do not have their effects because people believe they areinformative.

The power of random anchors has been demonstrated in someunsettling ways. German judges with an average of more than fifteen yearsof experience on the bench first read a description of a woman who hadbeen caught shoplifting, then rolled a pair of dice that were loaded soevery roll resulted in either a 3 or a 9. As soon as the dice came to a stop,the judges were asked whether they would sentence the woman to a termin prison greater or lesser, in months, than the number showing on thedice. Finally, the judges were instructed to specify the exact prisonsentence they would give to the shoplifter. On average, those who hadrolled a 9 said they would sentence her to 8 months; those who rolled a 3saidthif Africa they would sentence her to 5 months; the anchoring effectwas 50%.

Uses and Abuses of Anchors

By now you should be convinced that anchoring effects—sometimes dueto priming, sometimes to insufficient adjustment—are everywhere. Thepsychological mechanisms that produce anchoring make us far moresuggestible than most of us would want to be. And of course there arequite a few people who are willing and able to exploit our gullibility.

Anchoring effects explain why, for example, arbitrary rationing is aneffective marketing ploy. A few years ago, supermarket shoppers in SiouxCity, Iowa, encountered a sales promotion for Campbell’s soup at about10% off the regular price. On some days, a sign on the shelf said limit of12 per person. On other days, the sign said no limit per person. Shopperspurchased an average of 7 cans when the limit was in force, twice as manyas they bought when the limit was removed. Anchoring is not the sole

(PDF) Thinking, fast and slow by daniel kahneman - PDFSLIDE.US (125)

explanation. Rationing also implies that the goods are flying off theshelves, and shoppers should feel some urgency about stocking up. Butwe also know that the mention of 12 cans as a possible purchase wouldproduce anchoring even if the number were produced by a roulette wheel.

We see the same strategy at work in the negotiation over the price of ahome, when the seller makes the first move by setting the list price. As inmany other games, moving first is an advantage in single-issuenegotiations—for example, when price is the only issue to be settledbetween a buyer and a seller. As you may have experienced whennegotiating for the first time in a bazaar, the initial anchor has a powerfuleffect. My advice to students when I taught negotiations was that if youthink the other side has made an outrageous proposal, you should notcome back with an equally outrageous counteroffer, creating a gap that willbe difficult to bridge in further negotiations. Instead you should make ascene, storm out or threaten to do so, and make it clear—to yourself aswell as to the other side—that you will not continue the negotiation with thatnumber on the table.

The psychologists Adam Galinsky and Thomas Mussweiler proposedmore subtle ways to resist the anchoring effect in negotiations. Theyinstructed negotiators to focus their attention and search their memory forarguments against the anchor. The instruction to activate System 2 wassuccessful. For example, the anchoring effect is reduced or eliminatedwhen the second mover focuses his attention on the minimal offer that theopponent would accept, or on the costs to the opponent of failing to reachan agreement. In general, a strategy of deliberately “thinking the opposite”may be a good defense against anchoring effects, because it negates thebiased recruitment of thoughts that produces these effects.

Finally, try your hand at working out the effect of anchoring on a problemof public policy: the size of damages in personal injury cases. Theseawards are sometimes very large. Businesses that are frequent targets ofsuch lawsuits, such as hospitals and chemical companies, have lobbied toset a cap on the awards. Before you read this chapter you might havethought that capping awards is certainly good for potential defendants, butnow you should not be so sure. Consider the effect of capping awards at$1 million. This rule would eliminate all larger awards, but the anchor wouldalso pull up the size of many awards that would otherwise be much smaller.It would almost certainly benefit serious offenders and large firms muchmore than small ones.

Anchoring and the Two Systems

(PDF) Thinking, fast and slow by daniel kahneman - PDFSLIDE.US (126)

The effects of random anchors have much to tell us about the relationshipbetween System 1 and System 2. Anchoring effects have always beenstudied in tasks of judgment and choice that are ultimately completed bySystem 2. However, System 2 works on data that is retrieved frommemory, in an automatic and involuntary operation of System 1. System 2is therefore susceptible to the biasing influence of anchors that makesome information easier to retrieve. Furthermore, System 2 has no controlover the effect and no knowledge of it. The participants who have beenexposed to random or absurd anchors (such as Gandhi’s death at age144) confidently deny that this obviously useless information could haveinfluenced their estimate, and they are wrong.

We saw in the discussion of the law of small numbers that a message,unless it is immediately rejected as a lie, will have the same effect on theassociative system regardless of its reliability. The gist of the message isthe story, which is based on whatever information is available, even if thequantity of the information is slight and its quality is poor: WYSIATI. Whenyou read a story about the heroic rescue of a wounded mountain climber,its effect on your associative memory is much the same if it is a newsreport or the synopsis of a film. Anchoring results from this associativeactivation. Whether the story is true, or believable, matters little, if at all.The powerful effect of random anchors is an extreme case of thisphenomenon, because a random anchor obviously provides no informationat all.

Earlier I discussed the bewildering variety of priming effects, in whichyour thoughts and behavior may be influenced by stimuli to which you payno attention at all, and even by stimuli of which you are completelyunaware. The main moral of priming research is that our thoughts and ourbehavior are influenced, much more than we know or want, by theenvironment of the moment. Many people find the priming resultsunbelievable, because they do not correspond to subjective experience.Many others find the results upsetting, because they threaten the subjectivesense of agency and autonomy. If the content of a screen saver on anirrelevant computer can affect your willingness to help strangers withoutyour being aware of it, how free are you? Anchoring effects are threateningin a similar way. You are always aware of the anchor and even payattention to it, but you do not know how it guides and constrains yourthinking, because you cannot imagine how you would have thought if theanchor had been different (or absent). However, you should assume thatany number that is on the table has had an anchoring effect on you, and ifthe stakes are high you should mobilize yourself (your System 2) to combatthe effect.

(PDF) Thinking, fast and slow by daniel kahneman - PDFSLIDE.US (127)

Speaking of Anchors

“The firm we want to acquire sent us their business plan, with therevenue they expect. We shouldn’t let that number influence ourthinking. Set it aside.”

“Plans are best-case scenarios. Let’s avoid anchoring on planswhen we forecast actual outcomes. Thinking about ways the plancould go wrong is one way to do it.”

“Our aim in the negotiation is to get them anchored on thisnumber.”

& st

“The defendant’s lawyers put in a frivolous reference in which theymentioned a ridiculously low amount of damages, and they gotthe judge anchored on it!”

(PDF) Thinking, fast and slow by daniel kahneman - PDFSLIDE.US (128)

The Science of Availability

Amos and I had our most productive year in 1971–72, which we spent inEugene, Oregon. We were the guests of the Oregon Research Institute,which housed several future stars of all the fields in which we worked—judgment, decision making, and intuitive prediction. Our main host wasPaul Slovic, who had been Amos’s classmate at Ann Arbor and remaineda lifelong friend. Paul was on his way to becoming the leading psychologistamong scholars of risk, a position he has held for decades, collectingmany honors along the way. Paul and his wife, Roz, introduced us to life inEugene, and soon we were doing what people in Eugene do—jogging,barbecuing, and taking children to basketball games. We also worked veryhard, running dozens of experiments and writing our articles on judgmentheuristics. At night I wrote Attention and Effort. It was a busy year.

One of our projects was the study of what we called the availabilityheuristic. We thought of that heuristic when we asked ourselves whatpeople actually do when they wish to estimate the frequency of a category,such as “people who divorce after the age of 60” or “dangerous plants.”The answer was straightforward: instances of the class will be retrievedfrom memory, and if retrieval is easy and fluent, the category will be judgedto be large. We defined the availability heuristic as the process of judgingfrequency by “the ease with which instances come to mind.” The statementseemed clear when we formulated it, but the concept of availability hasbeen refined since then. The two-system approach had not yet beendeveloped when we studied availability, and we did not attempt todetermine whether this heuristic is a deliberate problem-solving strategy oran automatic operation. We now know that both systems are involved.

A question we considered early was how many instances must beretrieved to get an impression of the ease with which they come to mind.We now know the answer: none. For an example, think of the number ofwords that can be constructed from the two sets of letters below.

XUZONLCJMTAPCERHOB

You knew almost immediately, without generating any instances, that oneset offers far more possibilities than the other, probably by a factor of 10 ormore. Similarly, you do not need to retrieve specific news stories to have agood idea of the relative frequency with which different countries haveappeared in the news during the past year (Belgium, China, France,Congo, Nicaragua, Romania…).

(PDF) Thinking, fast and slow by daniel kahneman - PDFSLIDE.US (129)

The availability heuristic, like other heuristics of judgment, substitutesone question for another: you wish to estimate the size se ost c d of acategory or the frequency of an event, but you report an impression of theease with which instances come to mind. Substitution of questionsinevitably produces systematic errors. You can discover how the heuristicleads to biases by following a simple procedure: list factors other thanfrequency that make it easy to come up with instances. Each factor in yourlist will be a potential source of bias. Here are some examples:

A salient event that attracts your attention will be easily retrieved frommemory. Divorces among Hollywood celebrities and sex scandalsamong politicians attract much attention, and instances will comeeasily to mind. You are therefore likely to exaggerate the frequency ofboth Hollywood divorces and political sex scandals.A dramatic event temporarily increases the availability of itscategory. A plane crash that attracts media coverage will temporarilyalter your feelings about the safety of flying. Accidents are on yourmind, for a while, after you see a car burning at the side of the road,and the world is for a while a more dangerous place.Personal experiences, pictures, and vivid examples are moreavailable than incidents that happened to others, or mere words, orstatistics. A judicial error that affects you will undermine your faith inthe justice system more than a similar incident you read about in anewspaper.

Resisting this large collection of potential availability biases is possible,but tiresome. You must make the effort to reconsider your impressions andintuitions by asking such questions as, “Is our belief that theft s byteenagers are a major problem due to a few recent instances in ourneighborhood?” or “Could it be that I feel no need to get a flu shot becausenone of my acquaintances got the flu last year?” Maintaining one’svigilance against biases is a chore—but the chance to avoid a costlymistake is sometimes worth the effort.

One of the best-known studies of availability suggests that awareness ofyour own biases can contribute to peace in marriages, and probably inother joint projects. In a famous study, spouses were asked, “How largewas your personal contribution to keeping the place tidy, in percentages?”They also answered similar questions about “taking out the garbage,”“initiating social engagements,” etc. Would the self-estimated contributions

(PDF) Thinking, fast and slow by daniel kahneman - PDFSLIDE.US (130)

add up to 100%, or more, or less? As expected, the self-assessedcontributions added up to more than 100%. The explanation is a simpleavailability bias: both spouses remember their own individual efforts andcontributions much more clearly than those of the other, and the differencein availability leads to a difference in judged frequency. The bias is notnecessarily self-serving: spouses also overestimated their contribution tocausing quarrels, although to a smaller extent than their contributions tomore desirable outcomes. The same bias contributes to the commonobservation that many members of a collaborative team feel they havedone more than their share and also feel that the others are not adequatelygrateful for their individual contributions.

I am generally not optimistic about the potential for personal control ofbiases, but this is an exception. The opportunity for successful debiasingexists because the circumstances in which issues of credit allocationcome up are easy to identify, the more so because tensions often arisewhen several people at once feel that their efforts are not adequatelyrecognized. The mere observation that there is usually more than 100%credit to go around is sometimes sufficient to defuse the situation. In anyeve#82ght=nt, it is a good thing for every individual to remember. You willoccasionally do more than your share, but it is useful to know that you arelikely to have that feeling even when each member of the team feels thesame way.

The Psychology of Availability

A major advance in the understanding of the availability heuristic occurredin the early 1990s, when a group of German psychologists led by NorbertSchwarz raised an intriguing question: How will people’s impressions ofthe frequency of a category be affected by a requirement to list a specifiednumber of instances? Imagine yourself a subject in that experiment:

First, list six instances in which you behaved assertively.Next, evaluate how assertive you are.

Imagine that you had been asked for twelve instances of assertivebehavior (a number most people find difficult). Would your view of your ownassertiveness be different?

Schwarz and his colleagues observed that the task of listing instancesmay enhance the judgments of the trait by two different routes:

(PDF) Thinking, fast and slow by daniel kahneman - PDFSLIDE.US (131)

the number of instances retrievedthe ease with which they come to mind

The request to list twelve instances pits the two determinants against eachother. On the one hand, you have just retrieved an impressive number ofcases in which you were assertive. On the other hand, while the first threeor four instances of your own assertiveness probably came easily to you,you almost certainly struggled to come up with the last few to complete aset of twelve; fluency was low. Which will count more—the amount retrievedor the ease and fluency of the retrieval?

The contest yielded a clear-cut winner: people who had just listed twelveinstances rated themselves as less assertive than people who had listedonly six. Furthermore, participants who had been asked to list twelve casesin which they had not behaved assertively ended up thinking of themselvesas quite assertive! If you cannot easily come up with instances of meekbehavior, you are likely to conclude that you are not meek at all. Self-ratings were dominated by the ease with which examples had come tomind. The experience of fluent retrieval of instances trumped the numberretrieved.

An even more direct demonstration of the role of fluency was offered byother psychologists in the same group. All the participants in theirexperiment listed six instances of assertive (or nonassertive) behavior,while maintaining a specified facial expression. “Smilers” were instructedto contract the zygomaticus muscle, which produces a light smile;“frowners” were required to furrow their brow. As you already know,frowning normally accompanies cognitive strain and the effect issymmetric: when people are instructed to frown while doing a task, theyactually try harder and experience greater cognitive strain. Theresearchers anticipated that the frowners would have more difficultyretrieving examples of assertive behavior and would therefore ratethemselves as relatively lacking in assertiveness. And so it was.

Psychologists enjoy experiments that yield paradoxical results, and theyhave appliserv heighted Schwarz’s discovery with gusto. For example,people:

believe that they use their bicycles less often after recalling manyrather than few instances

(PDF) Thinking, fast and slow by daniel kahneman - PDFSLIDE.US (132)

are less confident in a choice when they are asked to produce morearguments to support itare less confident that an event was avoidable after listing moreways it could have been avoidedare less impressed by a car after listing many of its advantages

A professor at UCLA found an ingenious way to exploit the availabilitybias. He asked different groups of students to list ways to improve thecourse, and he varied the required number of improvements. As expected,the students who listed more ways to improve the class rated it higher!

Perhaps the most interesting finding of this paradoxical research is thatthe paradox is not always found: people sometimes go by content ratherthan by ease of retrieval. The proof that you truly understand a pattern ofbehavior is that you know how to reverse it. Schwarz and his colleaguestook on this challenge of discovering the conditions under which thisreversal would take place.

The ease with which instances of assertiveness come to the subject’smind changes during the task. The first few instances are easy, butretrieval soon becomes much harder. Of course, the subject also expectsfluency to drop gradually, but the drop of fluency between six and twelveinstances appears to be steeper than the participant expected. The resultssuggest that the participants make an inference: if I am having so muchmore trouble than expected coming up with instances of my assertiveness,then I can’t be very assertive. Note that this inference rests on a surprise—fluency being worse than expected. The availability heuristic that thesubjects apply is better described as an “unexplained unavailability”heuristic.

Schwarz and his colleagues reasoned that they could disrupt theheuristic by providing the subjects with an explanation for the fluency ofretrieval that they experienced. They told the participants they would hearbackground music while recalling instances and that the music would affectperformance in the memory task. Some subjects were told that the musicwould help, others were told to expect diminished fluency. As predicted,participants whose experience of fluency was “explained” did not use it asa heuristic; the subjects who were told that music would make retrievalmore difficult rated themselves as equally assertive when they retrievedtwelve instances as when they retrieved six. Other cover stories have beenused with the same result: judgments are no longer influenced by ease ofretrieval when the experience of fluency is given a spurious explanation bythe presence of curved or straight text boxes, by the background color ofthe screen, or by other irrelevant factors that the experimenters dreamed

(PDF) Thinking, fast and slow by daniel kahneman - PDFSLIDE.US (133)

up.As I have described it, the process that leads to judgment by availability

appears to involve a complex chain of reasoning. The subjects have anexperience of diminishing fluency as they produce instances. Theyevidently have expectations about the rate at which fluency decreases, andthose expectations are wrong: the difficulty of coming up with newinstances increases more rapidly than they expect. It is the unexpectedlylow fluency that causes people who were asked for twelve instances todescribe themselves as unassertive. When the surprise is eliminated, lowfluency no longer influences the judgment. The process appears to consistof a sophisticatedriethe subj set of inferences. Is the automatic System 1capable of it?

The answer is that in fact no complex reasoning is needed. Among thebasic features of System 1 is its ability to set expectations and to besurprised when these expectations are violated. The system also retrievespossible causes of a surprise, usually by finding a possible cause amongrecent surprises. Furthermore, System 2 can reset the expectations ofSystem 1 on the fly, so that an event that would normally be surprising isnow almost normal. Suppose you are told that the three-year-old boy wholives next door frequently wears a top hat in his stroller. You will be far lesssurprised when you actually see him with his top hat than you would havebeen without the warning. In Schwarz’s experiment, the background musichas been mentioned as a possible cause of retrieval problems. Thedifficulty of retrieving twelve instances is no longer a surprise and thereforeis less likely to be evoked by the task of judging assertiveness.

Schwarz and his colleagues discovered that people who are personallyinvolved in the judgment are more likely to consider the number ofinstances they retrieve from memory and less likely to go by fluency. Theyrecruited two groups of students for a study of risks to cardiac health. Halfthe students had a family history of cardiac disease and were expected totake the task more seriously than the others, who had no such history. Allwere asked to recall either three or eight behaviors in their routine thatcould affect their cardiac health (some were asked for risky behaviors,others for protective behaviors). Students with no family history of heartdisease were casual about the task and followed the availability heuristic.Students who found it difficult to find eight instances of risky behavior feltthemselves relatively safe, and those who struggled to retrieve examples ofsafe behaviors felt themselves at risk. The students with a family history ofheart disease showed the opposite pattern—they felt safer when theyretrieved many instances of safe behavior and felt greater danger whenthey retrieved many instances of risky behavior. They were also more likelyto feel that their future behavior would be affected by the experience of

(PDF) Thinking, fast and slow by daniel kahneman - PDFSLIDE.US (134)

evaluating their risk.The conclusion is that the ease with which instances come to mind is a

System 1 heuristic, which is replaced by a focus on content when System 2is more engaged. Multiple lines of evidence converge on the conclusionthat people who let themselves be guided by System 1 are more stronglysusceptible to availability biases than others who are in a state of highervigilance. The following are some conditions in which people “go with theflow” and are affected more strongly by ease of retrieval than by the contentthey retrieved:

when they are engaged in another effortful task at the same timewhen they are in a good mood because they just thought of a happyepisode in their lifeif they score low on a depression scaleif they are knowledgeable novices on the topic of the task, in contrastto true expertswhen they score high on a scale of faith in intuitionif they are (or are made to feel) powerful

I find the last finding particularly intriguing. The authors introduce theirarticle with a famous quote: “I don’t spend a lot of time taking polls aroundthe world to tell me what I think is the right way to act. I’ve just got to knowhow I feel” (Georgee e the w W. Bush, November 2002). They go on toshow that reliance on intuition is only in part a personality trait. Merelyreminding people of a time when they had power increases their apparenttrust in their own intuition.

Speaking of Availability

“Because of the coincidence of two planes crashing last month,she now prefers to take the train. That’s silly. The risk hasn’t reallychanged; it is an availability bias.”

“He underestimates the risks of indoor pollution because thereare few media stories on them. That’s an availability effect. Heshould look at the statistics.”

(PDF) Thinking, fast and slow by daniel kahneman - PDFSLIDE.US (135)

“She has been watching too many spy movies recently, so she’sseeing conspiracies everywhere.”

“The CEO has had several successes in a row, so failure doesn’tcome easily to her mind. The availability bias is making heroverconfident.”

(PDF) Thinking, fast and slow by daniel kahneman - PDFSLIDE.US (136)

Availability, Emotion, and Risk

Students of risk were quick to see that the idea of availability was relevantto their concerns. Even before our work was published, the economistHoward Kunreuther, who was then in the early stages of a career that hehas devoted to the study of risk and insurance, noticed that availabilityeffects help explain the pattern of insurance purchase and protective actionafter disasters. Victims and near victims are very concerned after adisaster. After each significant earthquake, Californians are for a whilediligent in purchasing insurance and adopting measures of protection andmitigation. They tie down their boiler to reduce quake damage, seal theirbasement doors against floods, and maintain emergency supplies in goodorder. However, the memories of the disaster dim over time, and so doworry and diligence. The dynamics of memory help explain the recurrentcycles of disaster, concern, and growing complacency that are familiar tostudents of large-scale emergencies.

Kunreuther also observed that protective actions, whether by individualsor governments, are usually designed to be adequate to the worst disasteractually experienced. As long ago as pharaonic Egypt, societies havetracked the high-water mark of rivers that periodically flood—and havealways prepared accordingly, apparently assuming that floods will not risehigher than the existing high-water mark. Images of a worse disaster donot come easily to mind.

Availability and Affect

The most influential studies of availability biases were carried out by ourfriends in Eugene, where Paul Slovic and his longtime collaborator SarahLichtenstein were joined by our former student Baruch Fischhoff. Theycarried out groundbreaking research on public perceptions of risks,including a survey that has become the standard example of an availabilitybias. They asked participants in their survey to siIs th t#consider pairs ofcauses of death: diabetes and asthma, or stroke and accidents. For eachpair, the subjects indicated the more frequent cause and estimated theratio of the two frequencies. The judgments were compared to healthstatistics of the time. Here’s a sample of their findings:

Strokes cause almost twice as many deaths as all accidentscombined, but 80% of respondents judged accidental death to be

(PDF) Thinking, fast and slow by daniel kahneman - PDFSLIDE.US (137)

more likely.Tornadoes were seen as more frequent killers than asthma, althoughthe latter cause 20 times more deaths.Death by lightning was judged less likely than death from botulismeven though it is 52 times more frequent.Death by disease is 18 times as likely as accidental death, but thetwo were judged about equally likely.Death by accidents was judged to be more than 300 times morelikely than death by diabetes, but the true ratio is 1:4.

The lesson is clear: estimates of causes of death are warped by mediacoverage. The coverage is itself biased toward novelty and poignancy. Themedia do not just shape what the public is interested in, but also areshaped by it. Editors cannot ignore the public’s demands that certaintopics and viewpoints receive extensive coverage. Unusual events (suchas botulism) attract disproportionate attention and are consequentlyperceived as less unusual than they really are. The world in our heads isnot a precise replica of reality; our expectations about the frequency ofevents are distorted by the prevalence and emotional intensity of themessages to which we are exposed.

The estimates of causes of death are an almost direct representation ofthe activation of ideas in associative memory, and are a good example ofsubstitution. But Slovic and his colleagues were led to a deeper insight:they saw that the ease with which ideas of various risks come to mind andthe emotional reactions to these risks are inextricably linked. Frighteningthoughts and images occur to us with particular ease, and thoughts ofdanger that are fluent and vivid exacerbate fear.

As mentioned earlier, Slovic eventually developed the notion of an affectheuristic, in which people make judgments and decisions by consultingtheir emotions: Do I like it? Do I hate it? How strongly do I feel about it? Inmany domains of life, Slovic said, people form opinions and make choicesthat directly express their feelings and their basic tendency to approach oravoid, often without knowing that they are doing so. The affect heuristic isan instance of substitution, in which the answer to an easy question (Howdo I feel about it?) serves as an answer to a much harder question (Whatdo I think about it?). Slovic and his colleagues related their views to thework of the neuroscientist Antonio Damasio, who had proposed thatpeople’s emotional evaluations of outcomes, and the bodily states and theapproach and avoidance tendencies associated with them, all play acentral role in guiding decision making. Damasio and his colleagues haveobserved that people who do not display the appropriate emotions before

(PDF) Thinking, fast and slow by daniel kahneman - PDFSLIDE.US (138)

they decide, sometimes because of brain damage, also have an impairedability to make good decisions. An inability to be guided by a “healthy fear”of bad consequences is a disastrous flaw.

In a compelling demonstration of the workings of the affect heuristic,Slovic’s research team surveyed opinions about various technologies,including water fluoridation, chemical plants, food preservatives, and cars,and asked their respondents to list both the benefits >

The best part of the experiment came next. After completing the initialsurvey, the respondents read brief passages with arguments in favor ofvarious technologies. Some were given arguments that focused on thenumerous benefits of a technology; others, arguments that stressed the lowrisks. These messages were effective in changing the emotional appeal ofthe technologies. The striking finding was that people who had received amessage extolling the benefits of a technology also changed their beliefsabout its risks. Although they had received no relevant evidence, thetechnology they now liked more than before was also perceived as lessrisky. Similarly, respondents who were told only that the risks of atechnology were mild developed a more favorable view of its benefits. Theimplication is clear: as the psychologist Jonathan Haidt said in anothercontext, “The emotional tail wags the rational dog.” The affect heuristicsimplifies our lives by creating a world that is much tidier than reality. Goodtechnologies have few costs in the imaginary world we inhabit, badtechnologies have no benefits, and all decisions are easy. In the real world,of course, we often face painful tradeoffs between benefits and costs.

The Public and the Experts

Paul Slovic probably knows more about the peculiarities of humanjudgment of risk than any other individual. His work offers a picture of Mr.and Ms. Citizen that is far from flattering: guided by emotion rather than byreason, easily swayed by trivial details, and inadequately sensitive todifferences between low and negligibly low probabilities. Slovic has alsostudied experts, who are clearly superior in dealing with numbers andamounts. Experts show many of the same biases as the rest of us inattenuated form, but often their judgments and preferences about risksdiverge from those of other people.

Differences between experts and the public are explained in part bybiases in lay judgments, but Slovic draws attention to situations in whichthe differences reflect a genuine conflict of values. He points out thatexperts often measure risks by the number of lives (or life-years) lost, whilethe public draws finer distinctions, for example between “good deaths” and

(PDF) Thinking, fast and slow by daniel kahneman - PDFSLIDE.US (139)

“bad deaths,” or between random accidental fatalities and deaths thatoccur in the course of voluntary activities such as skiing. These legitimatedistinctions are often ignored in statistics that merely count cases. Slovicargues from such observations that the public has a richer conception ofrisks than the experts do. Consequently, he strongly resists the view thatthe experts should rule, and that their opinions should be accepted withoutquestion when they conflict with the opinions and wishes of other citizens.When experts and the public disagree on their priorities, he says, “Eachside muiesst respect the insights and intelligence of the other.”

In his desire to wrest sole control of risk policy from experts, Slovic haschallenged the foundation of their expertise: the idea that risk is objective.

“Risk” does not exist “out there,” independent of our minds andculture, waiting to be measured. Human beings have invented theconcept of “risk” to help them understand and cope with thedangers and uncertainties of life. Although these dangers arereal, there is no such thing as “real risk” or “objective risk.”

To illustrate his claim, Slovic lists nine ways of defining the mortality riskassociated with the release of a toxic material into the air, ranging from“death per million people” to “death per million dollars of productproduced.” His point is that the evaluation of the risk depends on thechoice of a measure—with the obvious possibility that the choice mayhave been guided by a preference for one outcome or another. He goeson to conclude that “defining risk is thus an exercise in power.” You mightnot have guessed that one can get to such thorny policy issues fromexperimental studies of the psychology of judgment! However, policy isultimately about people, what they want and what is best for them. Everypolicy question involves assumptions about human nature, in particularabout the choices that people may make and the consequences of theirchoices for themselves and for society.

Another scholar and friend whom I greatly admire, Cass Sunstein,disagrees sharply with Slovic’s stance on the different views of experts andcitizens, and defends the role of experts as a bulwark against “populist”excesses. Sunstein is one of the foremost legal scholars in the UnitedStates, and shares with other leaders of his profession the attribute ofintellectual fearlessness. He knows he can master any body of knowledgequickly and thoroughly, and he has mastered many, including both thepsychology of judgment and choice and issues of regulation and riskpolicy. His view is that the existing system of regulation in the UnitedStates displays a very poor setting of priorities, which reflects reaction topublic pressures more than careful objective analysis. He starts from the

(PDF) Thinking, fast and slow by daniel kahneman - PDFSLIDE.US (140)

position that risk regulation and government intervention to reduce risksshould be guided by rational weighting of costs and benefits, and that thenatural units for this analysis are the number of lives saved (or perhaps thenumber of life-years saved, which gives more weight to saving the young)and the dollar cost to the economy. Poor regulation is wasteful of lives andmoney, both of which can be measured objectively. Sunstein has not beenpersuaded by Slovic’s argument that risk and its measurement issubjective. Many aspects of risk assessment are debatable, but he hasfaith in the objectivity that may be achieved by science, expertise, andcareful deliberation.

Sunstein came to believe that biased reactions to risks are an importantsource of erratic and misplaced priorities in public policy. Lawmakers andregulators may be overly responsive to the irrational concerns of citizens,both because of political sensitivity and because they are prone to thesame cognitive biases as other citizens.

Sunstein and a collaborator, the jurist Timur Kuran, invented a name forthe mechanism through which biases flow into policy: the availabilitycascade. They comment that in the social context, “all heuristics are equal,but availability is more equal than the others.” They have in mind an expandUned notion of the heuristic, in which availability provides a heuristic forjudgments other than frequency. In particular, the importance of an idea isoften judged by the fluency (and emotional charge) with which that ideacomes to mind.

An availability cascade is a self-sustaining chain of events, which maystart from media reports of a relatively minor event and lead up to publicpanic and large-scale government action. On some occasions, a mediastory about a risk catches the attention of a segment of the public, whichbecomes aroused and worried. This emotional reaction becomes a storyin itself, prompting additional coverage in the media, which in turnproduces greater concern and involvement. The cycle is sometimes spedalong deliberately by “availability entrepreneurs,” individuals ororganizations who work to ensure a continuous flow of worrying news. Thedanger is increasingly exaggerated as the media compete for attention-grabbing headlines. Scientists and others who try to dampen theincreasing fear and revulsion attract little attention, most of it hostile:anyone who claims that the danger is overstated is suspected ofassociation with a “heinous cover-up.” The issue becomes politicallyimportant because it is on everyone’s mind, and the response of thepolitical system is guided by the intensity of public sentiment. Theavailability cascade has now reset priorities. Other risks, and other waysthat resources could be applied for the public good, all have faded into the

(PDF) Thinking, fast and slow by daniel kahneman - PDFSLIDE.US (141)

background.Kuran and Sunstein focused on two examples that are still controversial:

the Love Canal affair and the so-called Alar scare. In Love Canal, buriedtoxic waste was exposed during a rainy season in 1979, causingcontamination of the water well beyond standard limits, as well as a foulsmell. The residents of the community were angry and frightened, and oneof them, Lois Gibbs, was particularly active in an attempt to sustain interestin the problem. The availability cascade unfolded according to thestandard script. At its peak there were daily stories about Love Canal,scientists attempting to claim that the dangers were overstated wereignored or shouted down, ABC News aired a program titled The KillingGround, and empty baby-size coffins were paraded in front of thelegislature. A large number of residents were relocated at governmentexpense, and the control of toxic waste became the major environmentalissue of the 1980s. The legislation that mandated the cleanup of toxicsites, called CERCLA, established a Superfund and is considered asignificant achievement of environmental legislation. It was also expensive,and some have claimed that the same amount of money could have savedmany more lives if it had been directed to other priorities. Opinions aboutwhat actually happened at Love Canal are still sharply divided, and claimsof actual damage to health appear not to have been substantiated. Kuranand Sunstein wrote up the Love Canal story almost as a pseudo-event,while on the other side of the debate, environmentalists still speak of the“Love Canal disaster.”

Opinions are also divided on the second example Kuran and Sunsteinused to illustrate their concept of an availability cascade, the Alar incident,known to detractors of environmental concerns as the “Alar scare” of 1989.Alar is a chemical that was sprayed on apples to regulate their growth andimprove their appearance. The scare began with press stories that thechemical, when consumed in gigantic doses, caused cancerous tumors inrats and mice. The stories understandably frightened the public, and thosefears encouraged more media coverage, the basic mechanism of anavailability cascade. The topic dominated the news and produceddramatic media events such as the testimony of the actress Meryl Streepbefore Congress. The apple industry su ofstained large losses as applesand apple products became objects of fear. Kuran and Sunstein quote acitizen who called in to ask “whether it was safer to pour apple juice downthe drain or to take it to a toxic waste dump.” The manufacturer withdrewthe product and the FDA banned it. Subsequent research confirmed thatthe substance might pose a very small risk as a possible carcinogen, butthe Alar incident was certainly an enormous overreaction to a minor

(PDF) Thinking, fast and slow by daniel kahneman - PDFSLIDE.US (142)

problem. The net effect of the incident on public health was probablydetrimental because fewer good apples were consumed.

The Alar tale illustrates a basic limitation in the ability of our mind to dealwith small risks: we either ignore them altogether or give them far too muchweight—nothing in between. Every parent who has stayed up waiting for ateenage daughter who is late from a party will recognize the feeling. Youmay know that there is really (almost) nothing to worry about, but youcannot help images of disaster from coming to mind. As Slovic hasargued, the amount of concern is not adequately sensitive to the probabilityof harm; you are imagining the numerator—the tragic story you saw on thenews—and not thinking about the denominator. Sunstein has coined thephrase “probability neglect” to describe the pattern. The combination ofprobability neglect with the social mechanisms of availability cascadesinevitably leads to gross exaggeration of minor threats, sometimes withimportant consequences.

In today’s world, terrorists are the most significant practitioners of the artof inducing availability cascades. With a few horrible exceptions such as9/11, the number of casualties from terror attacks is very small relative toother causes of death. Even in countries that have been targets ofintensive terror campaigns, such as Israel, the weekly number of casualtiesalmost never came close to the number of traffic deaths. The difference isin the availability of the two risks, the ease and the frequency with whichthey come to mind. Gruesome images, endlessly repeated in the media,cause everyone to be on edge. As I know from experience, it is difficult toreason oneself into a state of complete calm. Terrorism speaks directly toSystem 1.

Where do I come down in the debate between my friends? Availabilitycascades are real and they undoubtedly distort priorities in the allocationof public resources. Cass Sunstein would seek mechanisms that insulatedecision makers from public pressures, letting the allocation of resourcesbe determined by impartial experts who have a broad view of all risks andof the resources available to reduce them. Paul Slovic trusts the expertsmuch less and the public somewhat more than Sunstein does, and hepoints out that insulating the experts from the emotions of the publicproduces policies that the public will reject—an impossible situation in ademocracy. Both are eminently sensible, and I agree with both.

I share Sunstein’s discomfort with the influence of irrational fears andavailability cascades on public policy in the domain of risk. However, I alsoshare Slovic’s belief that widespread fears, even if they are unreasonable,should not be ignored by policy makers. Rational or not, fear is painful anddebilitating, and policy makers must endeavor to protect the public fromfear, not only from real dangers.

(PDF) Thinking, fast and slow by daniel kahneman - PDFSLIDE.US (143)

Slovic rightly stresses the resistance of the public to the idea ofdecisions being made by unelected and unaccountable experts.Furthermore, availability cascades may have a long-term benefit by callingattention to classes of risks and by increasing the overall size of the risk-reduction budget. The Love Canal incident may have caused excessiveresources to be allocated to the management of toxic betwaste, but it alsohad a more general effect in raising the priority level of environmentalconcerns. Democracy is inevitably messy, in part because the availabilityand affect heuristics that guide citizens’ beliefs and attitudes are inevitablybiased, even if they generally point in the right direction. Psychology shouldinform the design of risk policies that combine the experts’ knowledge withthe public’s emotions and intuitions.

Speaking of Availability Cascades

“She’s raving about an innovation that has large benefits and nocosts. I suspect the affect heuristic.”

“This is an availability cascade: a nonevent that is inflated by themedia and the public until it fills our TV screens and becomes allanyone is talking about.”

(PDF) Thinking, fast and slow by daniel kahneman - PDFSLIDE.US (144)

Tom W’s Specialty

Have a look at a simple puzzle:

Tom W is a graduate student at the main university in your state.Please rank the following nine fields of graduate specialization inorder of the likelihood that Tom W is now a student in each ofthese fields. Use 1 for the most likely, 9 for the least likely.

business administrationcomputer scienceengineeringhumanities and educationlawmedicinelibrary sciencephysical and life sciencessocial science and social work

This question is easy, and you knew immediately that the relative size ofenrollment in the different fields is the key to a solution. So far as you know,Tom W was picked at random from the graduate students at the university,like a single marble drawn from an urn. To decide whether a marble ismore likely to be red or green, you need to know how many marbles ofeach color there are in the urn. The proportion of marbles of a particularkind is called a base rate. Similarly, the base rate of humanities andeducation in this problem is the proportion of students of that field amongall the graduate students. In the absence of specific information about TomW, you will go by the base rates and guess that he is more likely to beenrolled in humanities and education than in computer science or libraryscience, because there are more students overall in the humanities andeducation than in the other two fields. Using base-rate information is theobvious move when no other information is provided.

Next comes a task that has nothing to do with base rates.

The following is a personality sketch of Tom W written duringTom’s senior year in high school by a psychologist, on the basisof psychological tests of uncertain validity:

(PDF) Thinking, fast and slow by daniel kahneman - PDFSLIDE.US (145)

Tom W is of high intelligence, although lacking in true creativity.He has a need for order and clarity, and for neat and tidy systemsin which every detail finds its appropriate place. His writing israther dull and mechanical, occasionally enlivened by somewhatcorny puns and flashes of imagination of the sci-fi type. He has astrong drive for competence. He seems to have little feel and littlesympathy for other people, and does not enjoy interacting withothers. Self-centered, he nonetheless has a deep moral sense.

Now please take a sheet of paper and rank the nine fields ofspecialization listed below by how similar the description of TomW is to the typical graduate student in each of the following fields.Use 1 for the most likely and 9 for the least likely.

You will get more out of the chapter if you give the task a quick try;reading the report on Tom W is necessary to make your judgments aboutthe various graduate specialties.

This question too is straightforward. It requires you to retrieve, orperhaps to construct, a stereotype of graduate students in the differentfields. When the experiment was first conducted, in the early 1970s, theaverage ordering was as follows. Yours is probably not very different:

1. computer science2. engineering3. business administration4. physical and life sciences5. library science6. law7. medicine8. humanities and education9. social science and social work

You probably ranked computer science among the best fitting because ofhints of nerdiness (“corny puns”). In fact, the description of Tom W waswritten to fit that stereotype. Another specialty that most people rankedhigh is engineering (“neat and tidy systems”). You probably thought thatTom W is not a good fit with your idea of social science and social work

(PDF) Thinking, fast and slow by daniel kahneman - PDFSLIDE.US (146)

(“little feel and little sympathy for other people”). Professional stereotypesappear to have changed little in the nearly forty years since I designed thedescription of Tom W.

The task of ranking the nine careers is complex and certainly requiresthe discipline and sequential organization of which only System 2 iscapable. However, the hints planted in the description (corny puns andothers) were intended to activate an association with a stereotype, anautomatic activity of System 1.

The instructions for this similarity task required a comparison of thedescription of Tom W to the stereotypes of the various fields ofspecialization. For the purposes of tv>

If you examine Tom W again, you will see that he is a good fit tostereotypes of some small groups of students (computer scientists,librarians, engineers) and a much poorer fit to the largest groups(humanities and education, social science and social work). Indeed, theparticipants almost always ranked the two largest fields very low. Tom Wwas intentionally designed as an “anti-base-rate” character, a good fit tosmall fields and a poor fit to the most populated specialties.

Predicting by Representativeness

The third task in the sequence was administered to graduate students inpsychology, and it is the critical one: rank the fields of specialization inorder of the likelihood that Tom W is now a graduate student in each ofthese fields. The members of this prediction group knew the relevantstatistical facts: they were familiar with the base rates of the different fields,and they knew that the source of Tom W’s description was not highlytrustworthy. However, we expected them to focus exclusively on thesimilarity of the description to the stereotypes—we called itrepresentativeness—ignoring both the base rates and the doubts aboutthe veracity of the description. They would then rank the small specialty—computer science—as highly probable, because that outcome gets thehighest representativeness score.

Amos and I worked hard during the year we spent in Eugene, and Isometimes stayed in the office through the night. One of my tasks for sucha night was to make up a description that would pit representativeness andbase rates against each other. Tom W was the result of my efforts, and Icompleted the description in the early morning hours. The first person whoshowed up to work that morning was our colleague and friend RobynDawes, who was both a sophisticated statistician and a skeptic about thevalidity of intuitive judgment. If anyone would see the relevance of the base

(PDF) Thinking, fast and slow by daniel kahneman - PDFSLIDE.US (147)

rate, it would have to be Robyn. I called Robyn over, gave him the questionI had just typed, and asked him to guess Tom W’s profession. I stillremember his sly smile as he said tentatively, “computer scientist?” Thatwas a happy moment—even the mighty had fallen. Of course, Robynimmediately recognized his mistake as soon as I mentioned “base rate,”but he had not spontaneously thought of it. Although he knew as much asanyone about the role of base rates in prediction, he neglected them whenpresented with the description of an individual’s personality. As expected,he substituted a judgment of representativeness for the probability he wasasked to assess.

Amos and I then collected answers to the same question from 114graduate students in psychology at three major universities, all of whomhad taken several courses in statistics. They did not disappoint us. Theirrankings of the nine fields by probability did not differ from ratings bysimilarity to the stereotype. Substitution was perfect in this case: there wasno indication that the participants did anything else but judgerepresentativeness. The question about probability (likelihood) wasdifficult, but the question about similarity was easier, and it was answeredinstead. This is a serious mistake, because judgments of similarity andprobak tbility are not constrained by the same logical rules. It is entirelyacceptable for judgments of similarity to be unaffected by base rates andalso by the possibility that the description was inaccurate, but anyone whoignores base rates and the quality of evidence in probability assessmentswill certainly make mistakes.

The concept “the probability that Tom W studies computer science” isnot a simple one. Logicians and statisticians disagree about its meaning,and some would say it has no meaning at all. For many experts it is ameasure of subjective degree of belief. There are some events you aresure of, for example, that the sun rose this morning, and others youconsider impossible, such as the Pacific Ocean freezing all at once. Thenthere are many events, such as your next-door neighbor being a computerscientist, to which you assign an intermediate degree of belief—which isyour probability of that event.

Logicians and statisticians have developed competing definitions ofprobability, all very precise. For laypeople, however, probability (asynonym of likelihood in everyday language) is a vague notion, related touncertainty, propensity, plausibility, and surprise. The vagueness is notparticular to this concept, nor is it especially troublesome. We know, moreor less, what we mean when we use a word such as democracy or beautyand the people we are talking to understand, more or less, what weintended to say. In all the years I spent asking questions about the

(PDF) Thinking, fast and slow by daniel kahneman - PDFSLIDE.US (148)

probability of events, no one ever raised a hand to ask me, “Sir, what doyou mean by probability?” as they would have done if I had asked them toassess a strange concept such as globability. Everyone acted as if theyknew how to answer my questions, although we all understood that it wouldbe unfair to ask them for an explanation of what the word means.

People who are asked to assess probability are not stumped, becausethey do not try to judge probability as statisticians and philosophers usethe word. A question about probability or likelihood activates a mentalshotgun, evoking answers to easier questions. One of the easy answers isan automatic assessment of representativeness—routine in understandinglanguage. The (false) statement that “Elvis Presley’s parents wanted him tobe a dentist” is mildly funny because the discrepancy between the imagesof Presley and a dentist is detected automatically. System 1 generates animpression of similarity without intending to do so. The representativenessheuristic is involved when someone says “She will win the election; you cansee she is a winner” or “He won’t go far as an academic; too manytattoos.” We rely on representativeness when we judge the potentialleadership of a candidate for office by the shape of his chin or theforcefulness of his speeches.

Although it is common, prediction by representativeness is notstatistically optimal. Michael Lewis’s bestselling Moneyball is a storyabout the inefficiency of this mode of prediction. Professional baseballscouts traditionally forecast the success of possible players in part by theirbuild and look. The hero of Lewis’s book is Billy Beane, the manager of theOakland A’s, who made the unpopular decision to overrule his scouts andto select players by the statistics of past performance. The players the A’spicked were inexpensive, because other teams had rejected them for notlooking the part. The team soon achieved excellent results at low cost.

The Sins of Representativeness

Judging probability byals representativeness has important virtues: theintuitive impressions that it produces are often—indeed, usually—moreaccurate than chance guesses would be.

On most occasions, people who act friendly are in fact friendly.A professional athlete who is very tall and thin is much more likely toplay basketball than football.People with a PhD are more likely to subscribe to The New YorkTimes than people who ended their education after high school.

(PDF) Thinking, fast and slow by daniel kahneman - PDFSLIDE.US (149)

Young men are more likely than elderly women to drive aggressively.

In all these cases and in many others, there is some truth to thestereotypes that govern judgments of representativeness, and predictionsthat follow this heuristic may be accurate. In other situations, thestereotypes are false and the representativeness heuristic will mislead,especially if it causes people to neglect base-rate information that points inanother direction. Even when the heuristic has some validity, exclusivereliance on it is associated with grave sins against statistical logic.

One sin of representativeness is an excessive willingness to predict theoccurrence of unlikely (low base-rate) events. Here is an example: you seea person reading The New York Times on the New York subway. Which ofthe following is a better bet about the reading stranger?

She has a PhD.She does not have a college degree.

Representativeness would tell you to bet on the PhD, but this is notnecessarily wise. You should seriously consider the second alternative,because many more nongraduates than PhDs ride in New York subways.And if you must guess whether a woman who is described as “a shy poetrylover” studies Chinese literature or business administration, you should optfor the latter option. Even if every female student of Chinese literature isshy and loves poetry, it is almost certain that there are more bashful poetrylovers in the much larger population of business students.

People without training in statistics are quite capable of using baserates in predictions under some conditions. In the first version of the TomW problem, which provides no details about him, it is obvious to everyonethat the probability of Tom W’s being in a particular field is simply the baserate frequency of enrollment in that field. However, concern for base ratesevidently disappears as soon as Tom W’s personality is described.

Amos and I originally believed, on the basis of our early evidence, thatbase-rate information will always be neglected when information about thespecific instance is available, but that conclusion was too strong.Psychologists have conducted many experiments in which base-rateinformation is explicitly provided as part of the problem, and many of theparticipants are influenced by those base rates, although the informationabout the individual case is almost always weighted more than merestatistics. Norbert Schwarz and his colleagues showed that instructingpeople to “think like a statistician” enhanced the use of base-rateinformation, while the instruction to “think like a clinician” had the opposite

(PDF) Thinking, fast and slow by daniel kahneman - PDFSLIDE.US (150)

effect.An experiment that was conducted a few years ago with Harvard

undergradut oates yielded a finding that surprised me: enhanced activationof System 2 caused a significant improvement of predictive accuracy inthe Tom W problem. The experiment combined the old problem with amodern variation of cognitive fluency. Half the students were told to puff outtheir cheeks during the task, while the others were told to frown. Frowning,as we have seen, generally increases the vigilance of System 2 andreduces both overconfidence and the reliance on intuition. The studentswho puffed out their cheeks (an emotionally neutral expression) replicatedthe original results: they relied exclusively on representativeness andignored the base rates. As the authors had predicted, however, thefrowners did show some sensitivity to the base rates. This is an instructivefinding.

When an incorrect intuitive judgment is made, System 1 and System 2should both be indicted. System 1 suggested the incorrect intuition, andSystem 2 endorsed it and expressed it in a judgment. However, there aretwo possible reasons for the failure of System 2—ignorance or laziness.Some people ignore base rates because they believe them to beirrelevant in the presence of individual information. Others make the samemistake because they are not focused on the task. If frowning makes adifference, laziness seems to be the proper explanation of base-rateneglect, at least among Harvard undergrads. Their System 2 “knows” thatbase rates are relevant even when they are not explicitly mentioned, butapplies that knowledge only when it invests special effort in the task.

The second sin of representativeness is insensitivity to the quality ofevidence. Recall the rule of System 1: WYSIATI. In the Tom W example,what activates your associative machinery is a description of Tom, whichmay or may not be an accurate portrayal. The statement that Tom W “haslittle feel and little sympathy for people” was probably enough to convinceyou (and most other readers) that he is very unlikely to be a student ofsocial science or social work. But you were explicitly told that thedescription should not be trusted!

You surely understand in principle that worthless information should notbe treated differently from a complete lack of information, but WY SIATImakes it very difficult to apply that principle. Unless you decideimmediately to reject evidence (for example, by determining that youreceived it from a liar), your System 1 will automatically process theinformation available as if it were true. There is one thing you can do whenyou have doubts about the quality of the evidence: let your judgments of

(PDF) Thinking, fast and slow by daniel kahneman - PDFSLIDE.US (151)

probability stay close to the base rate. Don’t expect this exercise ofdiscipline to be easy—it requires a significant effort of self-monitoring andself-control.

The correct answer to the Tom W puzzle is that you should stay veryclose to your prior beliefs, slightly reducing the initially high probabilities ofwell-populated fields (humanities and education; social science and socialwork) and slightly raising the low probabilities of rare specialties (libraryscience, computer science). You are not exactly where you would be if youhad known nothing at all about Tom W, but the little evidence you have isnot trustworthy, so the base rates should dominate your estimates.

How to Discipline Intuition

Your probability that it will rain tomorrow is your subjective degree of belief,but you should not let yourself believe whatever comes to your mind. To beuseful, your beliefs should be constrained by the logic of probability. So ifyou believe that there is a 40% chance plethat it will rain sometimetomorrow, you must also believe that there is a 60% chance it will not raintomorrow, and you must not believe that there is a 50% chance that it willrain tomorrow morning. And if you believe that there is a 30% chance thatcandidate X will be elected president, and an 80% chance that he will bereelected if he wins the first time, then you must believe that the chancesthat he will be elected twice in a row are 24%.

The relevant “rules” for cases such as the Tom W problem are providedby Bayesian statistics. This influential modern approach to statistics isnamed after an English minister of the eighteenth century, the ReverendThomas Bayes, who is credited with the first major contribution to a largeproblem: the logic of how people should change their mind in the light ofevidence. Bayes’s rule specifies how prior beliefs (in the examples of thischapter, base rates) should be combined with the diagnosticity of theevidence, the degree to which it favors the hypothesis over the alternative.For example, if you believe that 3% of graduate students are enrolled incomputer science (the base rate), and you also believe that the descriptionof Tom W is 4 times more likely for a graduate student in that field than inother fields, then Bayes’s rule says you must believe that the probabilitythat Tom W is a computer scientist is now 11%. If the base rate had been80%, the new degree of belief would be 94.1%. And so on.

The mathematical details are not relevant in this book. There are twoideas to keep in mind about Bayesian reasoning and how we tend to messit up. The first is that base rates matter, even in the presence of evidenceabout the case at hand. This is often not intuitively obvious. The second is

(PDF) Thinking, fast and slow by daniel kahneman - PDFSLIDE.US (152)

that intuitive impressions of the diagnosticity of evidence are oftenexaggerated. The combination of WY SIATI and associative coherencetends to make us believe in the stories we spin for ourselves. The essentialkeys to disciplined Bayesian reasoning can be simply summarized:

Anchor your judgment of the probability of an outcome on a plausiblebase rate.Question the diagnosticity of your evidence.

Both ideas are straightforward. It came as a shock to me when I realizedthat I was never taught how to implement them, and that even now I find itunnatural to do so.

Speaking of Representativeness

“The lawn is well trimmed, the receptionist looks competent, andthe furniture is attractive, but this doesn’t mean it is a well-managed company. I hope the board does not go byrepresentativeness.”

“This start-up looks as if it could not fail, but the base rate ofsuccess in the industry is extremely low. How do we know thiscase is different?”

“They keep making the same mistake: predicting rare eventsfrom weak evidence. When the evidence is weak, one shouldstick with the base rates.”

“I know this report is absolutely damning, and it may be based onsolid evidence, but how sure are we? We must allow for thatuncertainty in our thinking.”

ht="5%">

(PDF) Thinking, fast and slow by daniel kahneman - PDFSLIDE.US (153)

Linda: Less Is More

The best-known and most controversial of our experiments involved afictitious lady called Linda. Amos and I made up the Linda problem toprovide conclusive evidence of the role of heuristics in judgment and oftheir incompatibility with logic. This is how we described Linda:

Linda is thirty-one years old, single, outspoken, and very bright.She majored in philosophy. As a student, she was deeplyconcerned with issues of discrimination and social justice, andalso participated in antinuclear demonstrations.

The audiences who heard this description in the 1980s always laughedbecause they immediately knew that Linda had attended the University ofCalifornia at Berkeley, which was famous at the time for its radical,politically engaged students. In one of our experiments we presentedparticipants with a list of eight possible scenarios for Linda. As in the TomW problem, some ranked the scenarios by representativeness, others byprobability. The Linda problem is similar, but with a twist.

Linda is a teacher in elementary school.Linda works in a bookstore and takes yoga classes.Linda is active in the feminist movement.Linda is a psychiatric social worker.Linda is a member of the League of Women Voters.Linda is a bank teller.Linda is an insurance salesperson.Linda is a bank teller and is active in the feminist movement.

The problem shows its age in several ways. The League of Women Votersis no longer as prominent as it was, and the idea of a feminist “movement”sounds quaint, a testimonial to the change in the status of women over thelast thirty years. Even in the Facebook era, however, it is still easy to guessthe almost perfect consensus of judgments: Linda is a very good fit for anactive feminist, a fairly good fit for someone who works in a bookstore andtakes yoga classes—and a very poor fit for a bank teller or an insurancesalesperson.

Now focus on the critical items in the list: Does Linda look more like abank teller, or more like a bank teller who is active in the feministmovement? Everyone agrees that Linda fits the idea of a “feminist bankteller” better than she fits the stereotype of bank tellers. The stereotypicalbank teller is not a feminist activist, and adding that detail to the

(PDF) Thinking, fast and slow by daniel kahneman - PDFSLIDE.US (154)

bank teller is not a feminist activist, and adding that detail to thedescription makes for a more coherent story.

The twist comes in the judgments of likelihood, because there is alogical relation between the two scenarios. Think in terms of Venndiagrams. The set of feminist bank tellers is wholly included in the set ofbank tellers, as every feminist bank teller is0%"ustwora ban0%" w a bankteller. Therefore the probability that Linda is a feminist bank teller must belower than the probability of her being a bank teller. When you specify apossible event in greater detail you can only lower its probability. Theproblem therefore sets up a conflict between the intuition ofrepresentativeness and the logic of probability.

Our initial experiment was between-subjects. Each participant saw a setof seven outcomes that included only one of the critical items (“bank teller”or “feminist bank teller”). Some ranked the outcomes by resemblance,others by likelihood. As in the case of Tom W, the average rankings byresemblance and by likelihood were identical; “feminist bank teller” rankedhigher than “bank teller” in both.

Then we took the experiment further, using a within-subject design. Wemade up the questionnaire as you saw it, with “bank teller” in the sixthposition in the list and “feminist bank teller” as the last item. We wereconvinced that subjects would notice the relation between the twooutcomes, and that their rankings would be consistent with logic. Indeed,we were so certain of this that we did not think it worthwhile to conduct aspecial experiment. My assistant was running another experiment in thelab, and she asked the subjects to complete the new Linda questionnairewhile signing out, just before they got paid.

About ten questionnaires had accumulated in a tray on my assistant’sdesk before I casually glanced at them and found that all the subjects hadranked “feminist bank teller” as more probable than “bank teller.” I was sosurprised that I still retain a “flashbulb memory” of the gray color of themetal desk and of where everyone was when I made that discovery. Iquickly called Amos in great excitement to tell him what we had found: wehad pitted logic against representativeness, and representativeness hadwon!

In the language of this book, we had observed a failure of System 2: ourparticipants had a fair opportunity to detect the relevance of the logicalrule, since both outcomes were included in the same ranking. They did nottake advantage of that opportunity. When we extended the experiment, wefound that 89% of the undergraduates in our sample violated the logic ofprobability. We were convinced that statistically sophisticated respondentswould do better, so we administered the same questionnaire to doctoralstudents in the decision-science program of the Stanford Graduate School

(PDF) Thinking, fast and slow by daniel kahneman - PDFSLIDE.US (155)

of Business, all of whom had taken several advanced courses inprobability, statistics, and decision theory. We were surprised again: 85%of these respondents also ranked “feminist bank teller” as more likely than“bank teller.”

In what we later described as “increasingly desperate” attempts toeliminate the error, we introduced large groups of people to Linda andasked them this simple question:

Which alternative is more probable?Linda is a bank teller.Linda is a bank teller and is active in the feminist movement.

This stark version of the problem made Linda famous in some circles, andit earned us years of controversy. About 85% to 90% of undergraduates atseveral major universities chose the second option, contrary to logic.Remarkably, the sinners seemed to have no shame. When I asked mylarge undergraduatnite class in some indignation, “Do you realize that youhave violated an elementary logical rule?” someone in the back rowshouted, “So what?” and a graduate student who made the same errorexplained herself by saying, “I thought you just asked for my opinion.”

The word fallacy is used, in general, when people fail to apply a logicalrule that is obviously relevant. Amos and I introduced the idea of aconjunction fallacy, which people commit when they judge a conjunction oftwo events (here, bank teller and feminist) to be more probable than one ofthe events (bank teller) in a direct comparison.

As in the Müller-Lyer illusion, the fallacy remains attractive even whenyou recognize it for what it is. The naturalist Stephen Jay Gould describedhis own struggle with the Linda problem. He knew the correct answer, ofcourse, and yet, he wrote, “a little homunculus in my head continues to jumpup and down, shouting at me—‘but she can’t just be a bank teller; read thedescription.’” The little homunculus is of course Gould’s System 1speaking to him in insistent tones. (The two-system terminology had not yetbeen introduced when he wrote.)

The correct answer to the short version of the Linda problem was themajority response in only one of our studies: 64% of a group of graduatestudents in the social sciences at Stanford and at Berkeley correctlyjudged “feminist bank teller” to be less probable than “bank teller.” In theoriginal version with eight outcomes (shown above), only 15% of a similargroup of graduate students had made that choice. The difference isinstructive. The longer version separated the two critical outcomes by anintervening item (insurance salesperson), and the readers judged eachoutcome independently, without comparing them. The shorter version, in

(PDF) Thinking, fast and slow by daniel kahneman - PDFSLIDE.US (156)

contrast, required an explicit comparison that mobilized System 2 andallowed most of the statistically sophisticated students to avoid the fallacy.Unfortunately, we did not explore the reasoning of the substantial minority(36%) of this knowledgeable group who chose incorrectly.

The judgments of probability that our respondents offered, in both theTom W and Linda problems, corresponded precisely to judgments ofrepresentativeness (similarity to stereotypes). Representativenessbelongs to a cluster of closely related basic assessments that are likely tobe generated together. The most representative outcomes combine withthe personality description to produce the most coherent stories. The mostcoherent stories are not necessarily the most probable, but they areplausible, and the notions of coherence, plausibility, and probability areeasily confused by the unwary.

The uncritical substitution of plausibility for probability has perniciouseffects on judgments when scenarios are used as tools of forecasting.Consider these two scenarios, which were presented to different groups,with a request to evaluate their probability:

A massive flood somewhere in North America next year, in whichmore than 1,000 people drown

An earthquake in California sometime next year, causing a floodin which more than 1,000 people drown

The California earthquake scenario is more plausible than the NorthAmerica scenario, although its probability is certainly smaller. Asexpected, probability judgments were higher for the richer and moreentdetailed scenario, contrary to logic. This is a trap for forecasters andtheir clients: adding detail to scenarios makes them more persuasive, butless likely to come true.

To appreciate the role of plausibility, consider the following questions:

Which alternative is more probable?Mark has hair.Mark has blond hair.

and

Which alternative is more probable?Jane is a teacher.Jane is a teacher and walks to work.

(PDF) Thinking, fast and slow by daniel kahneman - PDFSLIDE.US (157)

The two questions have the same logical structure as the Linda problem,but they cause no fallacy, because the more detailed outcome is only moredetailed—it is not more plausible, or more coherent, or a better story. Theevaluation of plausibility and coherence does not suggest and answer tothe probability question. In the absence of a competing intuition, logicprevails.

Less Is More, Sometimes Even In Joint Evaluation

Christopher Hsee, of the University of Chicago, asked people to price setsof dinnerware offered in a clearance sale in a local store, wheredinnerware regularly runs between $30 and $60. There were three groupsin his experiment. The display below was shown to one group; Hsee labelsthat joint evaluation, because it allows a comparison of the two sets. Theother two groups were shown only one of the two sets; this is singleevaluation. Joint evaluation is a within-subject experiment, and singleevaluation is between-subjects.

Set A: 40 pieces Set B: 24 piecesDinner plates 8, all in good condition 8, all in good conditionSoup/salad bowls 8, all in good condition 8, all in good conditionDessert plates 8, all in good condition 8, all in good conditionCups 8, 2 of them brokenSaucers 8, 7 of them broken

Assuming that the dishes in the two sets are of equal quality, which isworth more? This question is easy. You can see that Set A contains all thedishes of Set B, and seven additional intact dishes, and it must be valuedmore. Indeed, the participants in Hsee’s joint evaluation experiment werewilling to pay a little more for Set A than for Set B: $32 versus $30.

The results reversed in single evaluation, where Set B was priced muchhigher than Set A: $33 versus $23. We know why this happened. Sets(including dinnerware sets!) are represented by norms and prototypes. Youcan sense immediately that the average value of the dishes is much lowerfor Set A than for Set B, because no one wants to pay for broken dishes. Ifthe average dominates the evaluation, it is not surprising that Set B isvalued more. Hsee called the resulting pattern less is more. By removing16 items from Set A (7 of them intact), its value is improved.

Hsee’s finding was replicated by the experimental economist John List

(PDF) Thinking, fast and slow by daniel kahneman - PDFSLIDE.US (158)

in a real market for baseball cards. He auctioned sets of ten high-valuecards, and identical sets to which three cards of modest value wereadded. As in the dinnerware experiment, the larger sets were valued morethan the smaller ones in joint evaluation, but less in single evaluation. Fromthe perspective of economic theory, this result is troubling: the economicvalue of a dinnerware set or of a collection of baseball cards is a sum-likevariable. Adding a positively valued item to the set can only increase itsvalue.

The Linda problem and the dinnerware problem have exactly the samestructure. Probability, like economic value, is a sum-like variable, asillustrated by this example:

probability (Linda is a teller) = probability (Linda is feminist teller)+ probability (Linda is non-feminist teller)

This is also why, as in Hsee’s dinnerware study, single evaluations of theLinda problem produce a less-is-more pattern. System 1 averages insteadof adding, so when the non-feminist bank tellers are removed from the set,subjective probability increases. However, the sum-like nature of thevariable is less obvious for probability than for money. As a result, jointevaluation eliminates the error only in Hsee’s experiment, not in the Lindaexperiment.

Linda was not the only conjunction error that survived joint evaluation.We found similar violations of logic in many other judgments. Participantsin one of these studies were asked to rank four possible outcomes of thenext Wimbledon tournament from most to least probable. Björn Borg wasthe dominant tennis player of the day when the study was conducted.These were the outcomes:

A. Borg will win the match.B. Borg will lose the first set.C. Borg will lose the first set but win the match.D. Borg will win the first set but lose the match.

The critical items are B and C. B is the more inclusive event and itsprobability must be higher than that of an event it includes. Contrary tologic, but not to representativeness or plausibility, 72% assigned B a lowerprobability than C—another instance of less is more in a directcomparison. Here si again, the scenario that was judged more probablewas unquestionably more plausible, a more coherent fit with all that wasknown about the best tennis player in the world.

To head off the possible objection that the conjunction fallacy is due to a

(PDF) Thinking, fast and slow by daniel kahneman - PDFSLIDE.US (159)

misinterpretation of probability, we constructed a problem that requiredprobability judgments, but in which the events were not described in words,and the term probability did not appear at all. We told participants about aregular six-sided die with four green faces and two red faces, which wouldbe rolled 20 times. They were shown three sequences of greens (G) andreds (R), and were asked to choose one. They would (hypothetically) win$25 if their chosen sequence showed up. The sequences were:

1. RGRRR2. GRGRRR3. GRRRRR

Because the die has twice as many green as red faces, the first sequenceis quite unrepresentative—like Linda being a bank teller. The secondsequence, which contains six tosses, is a better fit to what we wouldexpect from this die, because it includes two G’s. However, this sequencewas constructed by adding a G to the beginning of the first sequence, so itcan only be less likely than the first. This is the nonverbal equivalent toLinda being a feminist bank teller. As in the Linda study,representativeness dominated. Almost two-thirds of respondents preferredto bet on sequence 2 rather than on sequence 1. When presented witharguments for the two choices, however, a large majority found the correctargument (favoring sequence 1) more convincing.

The next problem was a breakthrough, because we finally found acondition in which the incidence of the conjunction fallacy was muchreduced. Two groups of subjects saw slightly different variants of the sameproblem:

(PDF) Thinking, fast and slow by daniel kahneman - PDFSLIDE.US (160)

The incidence of errors was 65% in the group that saw the problem on theleft, and only 25% in the group that saw the problem on the right.

Why is the question “How many of the 100 participants…” so mucheasier than “What percentage…”? A likely explanation is that the referenceto 100 individuals brings a spatial representation to mind. Imagine that alarge number of people are instructed to sort themselves into groups in aroom: “Those whose names begin with the letters A to L are told to gatherin the front left corner.” They are then instructed to sort themselves further.The relation of inclusion is now obvious, and you can see that individualswhose name begins with C will be a subset of the crowd in the front leftcorner. In the medical survey question, heart attack victims end up in acorner of the room, and some of them are less than 55 years old. Noteveryone will share this particular vivid imagery, but many subsequentexperiments have shown that the frequency representation, as it is known,makes it easy to appreciate that one group is wholly included in the other.The solution to the puzzle appears to be that a question phrased as “howmany?” makes you think of individuals, but the same question phrased as“what percentage?” does not.

What have we learned from these studies about the workings of System2? One conclusion, which is not new, is that System 2 is not impressivelyalert. The undergraduates and graduate students who participated in ourthastudies of the conjunction fallacy certainly “knew” the logic of Venndiagrams, but they did not apply it reliably even when all the relevantinformation was laid out in front of them. The absurdity of the less-is-morepattern was obvious in Hsee’s dinnerware study and was easilyrecognized in the “how many?” representation, but it was not apparent to

(PDF) Thinking, fast and slow by daniel kahneman - PDFSLIDE.US (161)

the thousands of people who have committed the conjunction fallacy in theoriginal Linda problem and in others like it. In all these cases, theconjunction appeared plausible, and that sufficed for an endorsement ofSystem 2.

The laziness of System 2 is part of the story. If their next vacation haddepended on it, and if they had been given indefinite time and told to followlogic and not to answer until they were sure of their answer, I believe thatmost of our subjects would have avoided the conjunction fallacy. However,their vacation did not depend on a correct answer; they spent very littletime on it, and were content to answer as if they had only been “asked fortheir opinion.” The laziness of System 2 is an important fact of life, and theobservation that representativeness can block the application of anobvious logical rule is also of some interest.

The remarkable aspect of the Linda story is the contrast to the broken-dishes study. The two problems have the same structure, but yield differentresults. People who see the dinnerware set that includes broken dishes puta very low price on it; their behavior reflects a rule of intuition. Others whosee both sets at once apply the logical rule that more dishes can only addvalue. Intuition governs judgments in the between-subjects condition; logicrules in joint evaluation. In the Linda problem, in contrast, intuition oftenovercame logic even in joint evaluation, although we identified someconditions in which logic prevails.

Amos and I believed that the blatant violations of the logic of probabilitythat we had observed in transparent problems were interesting and worthreporting to our colleagues. We also believed that the results strengthenedour argument about the power of judgment heuristics, and that they wouldpersuade doubters. And in this we were quite wrong. Instead, the Lindaproblem became a case study in the norms of controversy.

The Linda problem attracted a great deal of attention, but it also becamea magnet for critics of our approach to judgment. As we had already done,researchers found combinations of instructions and hints that reduced theincidence of the fallacy; some argued that, in the context of the Lindaproblem, it is reasonable for subjects to understand the word “probability”as if it means “plausibility.” These arguments were sometimes extended tosuggest that our entire enterprise was misguided: if one salient cognitiveillusion could be weakened or explained away, others could be as well.This reasoning neglects the unique feature of the conjunction fallacy as acase of conflict between intuition and logic. The evidence that we had builtup for heuristics from between-subjects experiment (including studies ofLinda) was not challenged—it was simply not addressed, and its saliencewas diminished by the exclusive focus on the conjunction fallacy. The neteffect of the Linda problem was an increase in the visibility of our work to

(PDF) Thinking, fast and slow by daniel kahneman - PDFSLIDE.US (162)

the general public, and a small dent in the credibility of our approachamong scholars in the field. This was not at all what we had expected.

If you visit a courtroom you will observe that lawyers apply two styles ofcriticism: to demolish a case they raise doubts about the strongestarguments that favor it; to discredit a witness, they focus on the weakestpart of the testimony. The focus on weaknesses is also normal inpoliticaverl debates. I do not believe it is appropriate in scientificcontroversies, but I have come to accept as a fact of life that the norms ofdebate in the social sciences do not prohibit the political style of argument,especially when large issues are at stake—and the prevalence of bias inhuman judgment is a large issue.

Some years ago I had a friendly conversation with Ralph Hertwig, apersistent critic of the Linda problem, with whom I had collaborated in avain attempt to settle our differences. I asked him why he and others hadchosen to focus exclusively on the conjunction fallacy, rather than on otherfindings that provided stronger support for our position. He smiled as heanswered, “It was more interesting,” adding that the Linda problem hadattracted so much attention that we had no reason to complain.

Speaking of Less is More

“They constructed a very complicated scenario and insisted oncalling it highly probable. It is not—it is only a plausible story.”

“They added a cheap gift to the expensive product, and made thewhole deal less attractive. Less is more in this case.”

“In most situations, a direct comparison makes people morecareful and more logical. But not always. Sometimes intuitionbeats logic even when the correct answer stares you in the face.”

(PDF) Thinking, fast and slow by daniel kahneman - PDFSLIDE.US (163)

Causes Trump Statistics

Consider the following scenario and note your intuitive answer to thequestion.

A cab was involved in a hit-and-run accident at night.Two cab companies, the Green and the Blue, operate in the city.You are given the following data:

85% of the cabs in the city are Green and 15% are Blue.A witness identified the cab as Blue. The court tested the reliability ofthe witness under the circumstances that existed on the night of theaccident and concluded that the witness correctly identified each oneof the two colors 80% of the time and failed 20% of the time.

What is the probability that the cab involved in the accident wasBlue rather than Green?

This is a standard problem of Bayesian inference. There are two items ofinformation: a base rate and the imperfectly reliable testimony of a witness.In the absence of a witness, the probability of the guilty cab being Blue is15%, which is the base rate of that outcome. If the two cab companies hadbeen equally large, the base rate would be uninformative and you wouldconsider only the reliability of the witness,%"> our w

Causal Stereotypes

Now consider a variation of the same story, in which only the presentationof the base rate has been altered.

You are given the following data:

The two companies operate the same number of cabs, but Greencabs are involved in 85% of accidents.

(PDF) Thinking, fast and slow by daniel kahneman - PDFSLIDE.US (164)

The information about the witness is as in the previous version.

The two versions of the problem are mathematically indistinguishable, butthey are psychologically quite different. People who read the first versiondo not know how to use the base rate and often ignore it. In contrast,people who see the second version give considerable weight to the baserate, and their average judgment is not too far from the Bayesian solution.Why?

In the first version, the base rate of Blue cabs is a statistical fact aboutthe cabs in the city. A mind that is hungry for causal stories finds nothing tochew on: How does the number of Green and Blue cabs in the city causethis cab driver to hit and run?

In the second version, in contrast, the drivers of Green cabs cause morethan 5 times as many accidents as the Blue cabs do. The conclusion isimmediate: the Green drivers must be a collection of reckless madmen!You have now formed a stereotype of Green recklessness, which you applyto unknown individual drivers in the company. The stereotype is easilyfitted into a causal story, because recklessness is a causally relevant factabout individual cabdrivers. In this version, there are two causal stories thatneed to be combined or reconciled. The first is the hit and run, whichnaturally evokes the idea that a reckless Green driver was responsible.The second is the witness’s testimony, which strongly suggests the cabwas Blue. The inferences from the two stories about the color of the car arecontradictory and approximately cancel each other. The chances for thetwo colors are about equal (the Bayesian estimate is 41%, reflecting thefact that the base rate of Green cabs is a little more extreme than thereliability of the witness who reported a Blue cab).

The cab example illustrates two types of base rates. Statistical baserates are facts about a population to which a case belongs, but they arenot relevant to the individual case. Causal base rates change your view ofhow the individual case came to be. The two types of base-rateinformation are treated differently:

Statistical base rates are generally underweighted, and sometimesneglected altogether, when specific information about the case athand is available.Causal base rates are treated as information about the individualcase and are easily combined with other case-specific information.

(PDF) Thinking, fast and slow by daniel kahneman - PDFSLIDE.US (165)

The causal version of the cab problem had the form of a stereotype: Greendrivers are dangerous. Stereotypes are statements about the group thatare (at least tentatively) accepted as facts about every member. Hely reare two examples:

Most of the graduates of this inner-city school go to college.Interest in cycling is widespread in France.

These statements are readily interpreted as setting up a propensity inindividual members of the group, and they fit in a causal story. Manygraduates of this particular inner-city school are eager and able to go tocollege, presumably because of some beneficial features of life in thatschool. There are forces in French culture and social life that cause manyFrenchmen to take an interest in cycling. You will be reminded of thesefacts when you think about the likelihood that a particular graduate of theschool will attend college, or when you wonder whether to bring up the Tourde France in a conversation with a Frenchman you just met.

Stereotyping is a bad word in our culture, but in my usage it is neutral. Oneof the basic characteristics of System 1 is that it represents categories asnorms and prototypical exemplars. This is how we think of horses,refrigerators, and New York police officers; we hold in memory arepresentation of one or more “normal” members of each of thesecategories. When the categories are social, these representations arecalled stereotypes. Some stereotypes are perniciously wrong, and hostilestereotyping can have dreadful consequences, but the psychological factscannot be avoided: stereotypes, both correct and false, are how we thinkof categories.

You may note the irony. In the context of the cab problem, the neglect ofbase-rate information is a cognitive flaw, a failure of Bayesian reasoning,and the reliance on causal base rates is desirable. Stereotyping the Greendrivers improves the accuracy of judgment. In other contexts, however,such as hiring or profiling, there is a strong social norm againststereotyping, which is also embedded in the law. This is as it should be. Insensitive social contexts, we do not want to draw possibly erroneousconclusions about the individual from the statistics of the group. Weconsider it morally desirable for base rates to be treated as statistical factsabout the group rather than as presumptive facts about individuals. In otherwords, we reject causal base rates.

The social norm against stereotyping, including the opposition to

(PDF) Thinking, fast and slow by daniel kahneman - PDFSLIDE.US (166)

profiling, has been highly beneficial in creating a more civilized and moreequal society. It is useful to remember, however, that neglecting validstereotypes inevitably results in suboptimal judgments. Resistance tostereotyping is a laudable moral position, but the simplistic idea that theresistance is costless is wrong. The costs are worth paying to achieve abetter society, but denying that the costs exist, while satisfying to the souland politically correct, is not scientifically defensible. Reliance on the affectheuristic is common in politically charged arguments. The positions wefavor have no cost and those we oppose have no benefits. We should beable to do better.

Causal Situations

Amos and I constructed the variants of the cab problem, but we did notinvent the powerful notion of causal base rates; we borrowed it from thepsychologist Icek Ajzen. In his experiment, Ajzen showed his participantsbrief vignettes describing some students who had taken an exam at Yaleand asked the participants to judge the probability that each student hadpassed the test. The manipulation of causal bs oase rates wasstraightforward: Ajzen told one group that the students they saw had beendrawn from a class in which 75% passed the exam, and told another groupthat the same students had been in a class in which only 25% passed. Thisis a powerful manipulation, because the base rate of passing suggests theimmediate inference that the test that only 25% passed must have beenbrutally difficult. The difficulty of a test is, of course, one of the causalfactors that determine every student’s outcome. As expected, Ajzen’ssubjects were highly sensitive to the causal base rates, and every studentwas judged more likely to pass in the high-success condition than in thehigh-failure rate.

Ajzen used an ingenious method to suggest a noncausal base rate. Hetold his subjects that the students they saw had been drawn from a sample,which itself was constructed by selecting students who had passed orfailed the exam. For example, the information for the high-failure groupread as follows:

The investigator was mainly interested in the causes of failureand constructed a sample in which 75% had failed theexamination.

Note the difference. This base rate is a purely statistical fact about theensemble from which cases have been drawn. It has no bearing on thequestion asked, which is whether the individual student passed or failed

(PDF) Thinking, fast and slow by daniel kahneman - PDFSLIDE.US (167)

the test. As expected, the explicitly stated base rates had some effects onjudgment, but they had much less impact than the statistically equivalentcausal base rates. System 1 can deal with stories in which the elementsare causally linked, but it is weak in statistical reasoning. For a Bayesianthinker, of course, the versions are equivalent. It is tempting to concludethat we have reached a satisfactory conclusion: causal base rates areused; merely statistical facts are (more or less) neglected. The next study,one of my all-time favorites, shows that the situation is rather morecomplex.

Can Psychology be Taught?

The reckless cabdrivers and the impossibly difficult exam illustrate twoinferences that people can draw from causal base rates: a stereotypicaltrait that is attributed to an individual, and a significant feature of thesituation that affects an individual’s outcome. The participants in theexperiments made the correct inferences and their judgments improved.Unfortunately, things do not always work out so well. The classicexperiment I describe next shows that people will not draw from base-rateinformation an inference that conflicts with other beliefs. It also supports theuncomfortable conclusion that teaching psychology is mostly a waste oftime.

The experiment was conducted a long time ago by the socialpsychologist Richard Nisbett and his student Eugene Borgida, at theUniversity of Michigan. They told students about the renowned “helpingexperiment” that had been conducted a few years earlier at New YorkUniversity. Participants in that experiment were led to individual boothsand invited to speak over the intercom about their personal lives andproblems. They were to talk in turn for about two minutes. Only onemicrophone was active at any one time. There were six participants ineach group, one of whom was a stooge. The stooge spoke first, followinga script prepared by the experimenters. He described his problemsadjusting to New York and admitted with obvious embarrassment that hewas prone to seizures, especially when stressed. All the participants thenhad a turn. When the microphone was again turned over to the stooge, hebecame agitated and incoherent, said he felt a seizure coming on, andpeoasked for someone to help him. The last words heard from him were, “C-could somebody-er-er-help-er-uh-uh-uh [choking sounds]. I…I’m gonna die-er-er-er I’m…gonna die-er-er-I seizure I-er [chokes, then quiet].” At thispoint the microphone of the next participant automatically became active,and nothing more was heard from the possibly dying individual.

(PDF) Thinking, fast and slow by daniel kahneman - PDFSLIDE.US (168)

What do you think the participants in the experiment did? So far as theparticipants knew, one of them was having a seizure and had asked forhelp. However, there were several other people who could possiblyrespond, so perhaps one could stay safely in one’s booth. These were theresults: only four of the fifteen participants responded immediately to theappeal for help. Six never got out of their booth, and five others came outonly well after the “seizure victim” apparently choked. The experimentshows that individuals feel relieved of responsibility when they know thatothers have heard the same request for help.

Did the results surprise you? Very probably. Most of us think ofourselves as decent people who would rush to help in such a situation, andwe expect other decent people to do the same. The point of theexperiment, of course, was to show that this expectation is wrong. Evennormal, decent people do not rush to help when they expect others to takeon the unpleasantness of dealing with a seizure. And that means you, too.

Are you willing to endorse the following statement? “When I read theprocedure of the helping experiment I thought I would come to thestranger’s help immediately, as I probably would if I found myself alone witha seizure victim. I was probably wrong. If I find myself in a situation in whichother people have an opportunity to help, I might not step forward. Thepresence of others would reduce my sense of personal responsibility morethan I initially thought.” This is what a teacher of psychology would hope youwould learn. Would you have made the same inferences by yourself?

The psychology professor who describes the helping experiment wantsthe students to view the low base rate as causal, just as in the case of thefictitious Yale exam. He wants them to infer, in both cases, that asurprisingly high rate of failure implies a very difficult test. The lessonstudents are meant to take away is that some potent feature of thesituation, such as the diffusion of responsibility, induces normal and decentpeople such as them to behave in a surprisingly unhelpful way.

Changing one’s mind about human nature is hard work, and changingone’s mind for the worse about oneself is even harder. Nisbett andBorgida suspected that students would resist the work and theunpleasantness. Of course, the students would be able and willing to recitethe details of the helping experiment on a test, and would even repeat the“official” interpretation in terms of diffusion of responsibility. But did theirbeliefs about human nature really change? To find out, Nisbett and Borgidashowed them videos of brief interviews allegedly conducted with twopeople who had participated in the New York study. The interviews wereshort and bland. The interviewees appeared to be nice, normal, decentpeople. They described their hobbies, their spare-time activities, and theirplans for the future, which were entirely conventional. After watching the

(PDF) Thinking, fast and slow by daniel kahneman - PDFSLIDE.US (169)

video of an interview, the students guessed how quickly that particularperson had come to the aid of the stricken stranger.

To apply Bayesian reasoning to the task the students were assigned, youshould first ask yourself what you would have guessed about the a stwoindividuals if you had not seen their interviews. This question is answeredby consulting the base rate. We have been told that only 4 of the 15participants in the experiment rushed to help after the first request. Theprobability that an unidentified participant had been immediately helpful istherefore 27%. Thus your prior belief about any unspecified participantshould be that he did not rush to help. Next, Bayesian logic requires you toadjust your judgment in light of any relevant information about theindividual. However, the videos were carefully designed to beuninformative; they provided no reason to suspect that the individualswould be either more or less helpful than a randomly chosen student. In theabsence of useful new information, the Bayesian solution is to stay with thebase rates.

Nisbett and Borgida asked two groups of students to watch the videosand predict the behavior of the two individuals. The students in the firstgroup were told only about the procedure of the helping experiment, notabout its results. Their predictions reflected their views of human natureand their understanding of the situation. As you might expect, theypredicted that both individuals would immediately rush to the victim’s aid.The second group of students knew both the procedure of the experimentand its results. The comparison of the predictions of the two groupsprovides an answer to a significant question: Did students learn from theresults of the helping experiment anything that significantly changed theirway of thinking? The answer is straightforward: they learned nothing at all.Their predictions about the two individuals were indistinguishable from thepredictions made by students who had not been exposed to the statisticalresults of the experiment. They knew the base rate in the group from whichthe individuals had been drawn, but they remained convinced that thepeople they saw on the video had been quick to help the stricken stranger.

For teachers of psychology, the implications of this study aredisheartening. When we teach our students about the behavior of people inthe helping experiment, we expect them to learn something they had notknown before; we wish to change how they think about people’s behaviorin a particular situation. This goal was not accomplished in the Nisbett-Borgida study, and there is no reason to believe that the results would havebeen different if they had chosen another surprising psychological

(PDF) Thinking, fast and slow by daniel kahneman - PDFSLIDE.US (170)

experiment. Indeed, Nisbett and Borgida reported similar findings inteaching another study, in which mild social pressure caused people toaccept much more painful electric shocks than most of us (and them)would have expected. Students who do not develop a new appreciation forthe power of social setting have learned nothing of value from theexperiment. The predictions they make about random strangers, or abouttheir own behavior, indicate that they have not changed their view of howthey would have behaved. In the words of Nisbett and Borgida, students“quietly exempt themselves” (and their friends and acquaintances) from theconclusions of experiments that surprise them. Teachers of psychologyshould not despair, however, because Nisbett and Borgida report a way tomake their students appreciate the point of the helping experiment. Theytook a new group of students and taught them the procedure of theexperiment but did not tell them the group results. They showed the twovideos and simply told their students that the two individuals they had justseen had not helped the stranger, then asked them to guess the globalresults. The outcome was dramatic: the students’ guesses were extremelyaccurate.

To teach students any psychology they did not know before, you mustsurprise them. But which surprise will do? Nisbett and Borgida found thatwhen they presented their students with a surprising statisticis al fact, thestudents managed to learn nothing at all. But when the students weresurprised by individual cases—two nice people who had not helped—theyimmediately made the generalization and inferred that helping is moredifficult than they had thought. Nisbett and Borgida summarize the resultsin a memorable sentence:

Subjects’ unwillingness to deduce the particular from the generalwas matched only by their willingness to infer the general from theparticular.

This is a profoundly important conclusion. People who are taughtsurprising statistical facts about human behavior may be impressed to thepoint of telling their friends about what they have heard, but this does notmean that their understanding of the world has really changed. The test oflearning psychology is whether your understanding of situations youencounter has changed, not whether you have learned a new fact. There isa deep gap between our thinking about statistics and our thinking aboutindividual cases. Statistical results with a causal interpretation have astronger effect on our thinking than noncausal information. But evencompelling causal statistics will not change long-held beliefs or beliefsrooted in personal experience. On the other hand, surprising individual

(PDF) Thinking, fast and slow by daniel kahneman - PDFSLIDE.US (171)

cases have a powerful impact and are a more effective tool for teachingpsychology because the incongruity must be resolved and embedded in acausal story. That is why this book contains questions that are addressedpersonally to the reader. You are more likely to learn something by findingsurprises in your own behavior than by hearing surprising facts aboutpeople in general.

Speaking of Causes and Statistics

“We can’t assume that they will really learn anything from merestatistics. Let’s show them one or two representative individualcases to influence their System 1.”

“No need to worry about this statistical information being ignored.On the contrary, it will immediately be used to feed a stereotype.”

(PDF) Thinking, fast and slow by daniel kahneman - PDFSLIDE.US (172)

Regression to the Mean

I had one of the most satisfying eureka experiences of my career whileteaching flight instructors in the Israeli Air Force about the psychology ofeffective training. I was telling them about an important principle of skilltraining: rewards for improved performance work better than punishment ofmistakes. This proposition is supported by much evidence from researchon pigeons, rats, humans, and other animals.

When I finished my enthusiastic speech, one of the most seasonedinstructors in the group raised his hand and made a short speech of hisown. He began by conceding that rewarding improved performance mightbe good for the birds, but he denied that it was optimal for flight cadets.This is what he said: “On many occasions I have praised flight cadets forclean execution of some aerobatic maneuver. The next time they try thesame maneuver they usually do worse. On the other hand, I have oftenscreamed into a cadet’s earphone for bad execution, and in general hedoes better t t ask yry abr two repon his next try. So please don’t tell us thatreward works and punishment does not, because the opposite is thecase.”

This was a joyous moment of insight, when I saw in a new light aprinciple of statistics that I had been teaching for years. The instructor wasright—but he was also completely wrong! His observation was astute andcorrect: occasions on which he praised a performance were likely to befollowed by a disappointing performance, and punishments were typicallyfollowed by an improvement. But the inference he had drawn about theefficacy of reward and punishment was completely off the mark. What hehad observed is known as regression to the mean, which in that case wasdue to random fluctuations in the quality of performance. Naturally, hepraised only a cadet whose performance was far better than average. Butthe cadet was probably just lucky on that particular attempt and thereforelikely to deteriorate regardless of whether or not he was praised. Similarly,the instructor would shout into a cadet’s earphones only when the cadet’sperformance was unusually bad and therefore likely to improve regardlessof what the instructor did. The instructor had attached a causalinterpretation to the inevitable fluctuations of a random process.

The challenge called for a response, but a lesson in the algebra ofprediction would not be enthusiastically received. Instead, I used chalk tomark a target on the floor. I asked every officer in the room to turn his backto the target and throw two coins at it in immediate succession, withoutlooking. We measured the distances from the target and wrote the tworesults of each contestant on the blackboard. Then we rewrote the results

(PDF) Thinking, fast and slow by daniel kahneman - PDFSLIDE.US (173)

in order, from the best to the worst performance on the first try. It wasapparent that most (but not all) of those who had done best the first timedeteriorated on their second try, and those who had done poorly on the firstattempt generally improved. I pointed out to the instructors that what theysaw on the board coincided with what we had heard about theperformance of aerobatic maneuvers on successive attempts: poorperformance was typically followed by improvement and goodperformance by deterioration, without any help from either praise orpunishment.

The discovery I made on that day was that the flight instructors weretrapped in an unfortunate contingency: because they punished cadetswhen performance was poor, they were mostly rewarded by a subsequentimprovement, even if punishment was actually ineffective. Furthermore, theinstructors were not alone in that predicament. I had stumbled onto asignificant fact of the human condition: the feedback to which life exposesus is perverse. Because we tend to be nice to other people when theyplease us and nasty when they do not, we are statistically punished forbeing nice and rewarded for being nasty.

Talent and Luck

A few years ago, John Brockman, who edits the online magazine Edge,asked a number of scientists to report their “favorite equation.” These weremy offerings:

success = talent + luckgreat success = a little more talent + a lot of luck

The unsurprising idea that luck often contributes to success has surprisingconsequences when we apply it to the first two days of a high-level golftournament. To keep things simple, assume that on both days the averagescore of the competitors was at par 72. We focus on a player who didverye d well on the first day, closing with a score of 66. What can we learnfrom that excellent score? An immediate inference is that the golfer ismore talented than the average participant in the tournament. The formulafor success suggests that another inference is equally justified: the golferwho did so well on day 1 probably enjoyed better-than-average luck on thatday. If you accept that talent and luck both contribute to success, theconclusion that the successful golfer was lucky is as warranted as theconclusion that he is talented.

By the same token, if you focus on a player who scored 5 over par on

(PDF) Thinking, fast and slow by daniel kahneman - PDFSLIDE.US (174)

that day, you have reason to infer both that he is rather weak and had abad day. Of course, you know that neither of these inferences is certain. Itis entirely possible that the player who scored 77 is actually very talentedbut had an exceptionally dreadful day. Uncertain though they are, thefollowing inferences from the score on day 1 are plausible and will becorrect more often than they are wrong.

above-average score on day 1 = above-average talent + lucky onday 1

and

below-average score on day 1 = below-average talent + unluckyon day 1

Now, suppose you know a golfer’s score on day 1 and are asked topredict his score on day 2. You expect the golfer to retain the same level oftalent on the second day, so your best guesses will be “above average” forthe first player and “below average” for the second player. Luck, of course,is a different matter. Since you have no way of predicting the golfers’ luckon the second (or any) day, your best guess must be that it will be average,neither good nor bad. This means that in the absence of any otherinformation, your best guess about the players’ score on day 2 should notbe a repeat of their performance on day 1. This is the most you can say:

The golfer who did well on day 1 is likely to be successful on day 2 aswell, but less than on the first, because the unusual luck he probablyenjoyed on day 1 is unlikely to hold.The golfer who did poorly on day 1 will probably be below averageon day 2, but will improve, because his probable streak of bad luck isnot likely to continue.

We also expect the difference between the two golfers to shrink on thesecond day, although our best guess is that the first player will still dobetter than the second.

My students were always surprised to hear that the best predictedperformance on day 2 is more moderate, closer to the average than theevidence on which it is based (the score on day 1). This is why the patternis called regression to the mean. The more extreme the original score, the

(PDF) Thinking, fast and slow by daniel kahneman - PDFSLIDE.US (175)

more regression we expect, because an extremely good score suggests avery lucky day. The regressive prediction is reasonable, but its accuracy isnot guaranteed. A few of the golfers who scored 66 on day 1 will do evenbetter on the second day, if their luck improves. Most will do worse,because their luck will no longer be above average.

Now let us go against the time arrow. Arrange the players by theirperformance on day 2 and look at their performance on day 1. You will findprecisely the same pattern of regression to the mean. The golfers who didbest on day 2 were probably lucky on that day, and the best guess is thatthey had been less lucky and had done filess well on day 1. The fact thatyou observe regression when you predict an early event from a later eventshould help convince you that regression does not have a causalexplanation.

Regression effects are ubiquitous, and so are misguided causal storiesto explain them. A well-known example is the “Sports Illustrated jinx,” theclaim that an athlete whose picture appears on the cover of the magazineis doomed to perform poorly the following season. Overconfidence and thepressure of meeting high expectations are often offered as explanations.But there is a simpler account of the jinx: an athlete who gets to be on thecover of Sports Illustrated must have performed exceptionally well in thepreceding season, probably with the assistance of a nudge from luck—andluck is fickle.

I happened to watch the men’s ski jump event in the Winter Olympicswhile Amos and I were writing an article about intuitive prediction. Eachathlete has two jumps in the event, and the results are combined for thefinal score. I was startled to hear the sportscaster’s comments whileathletes were preparing for their second jump: “Norway had a great firstjump; he will be tense, hoping to protect his lead and will probably doworse” or “Sweden had a bad first jump and now he knows he has nothingto lose and will be relaxed, which should help him do better.” Thecommentator had obviously detected regression to the mean and hadinvented a causal story for which there was no evidence. The story itselfcould even be true. Perhaps if we measured the athletes’ pulse beforeeach jump we might find that they are indeed more relaxed after a bad firstjump. And perhaps not. The point to remember is that the change from thefirst to the second jump does not need a causal explanation. It is amathematically inevitable consequence of the fact that luck played a role inthe outcome of the first jump. Not a very satisfactory story—we would allprefer a causal account—but that is all there is.

Understanding Regression

(PDF) Thinking, fast and slow by daniel kahneman - PDFSLIDE.US (176)

Whether undetected or wrongly explained, the phenomenon of regressionis strange to the human mind. So strange, indeed, that it was first identifiedand understood two hundred years after the theory of gravitation anddifferential calculus. Furthermore, it took one of the best minds ofnineteenth-century Britain to make sense of it, and that with great difficulty.

Regression to the mean was discovered and named late in thenineteenth century by Sir Francis Galton, a half cousin of Charles Darwinand a renowned polymath. You can sense the thrill of discovery in an articlehe published in 1886 under the title “Regression towards Mediocrity inHereditary Stature,” which reports measurements of size in successivegenerations of seeds and in comparisons of the height of children to theheight of their parents. He writes about his studies of seeds:

They yielded results that seemed very noteworthy, and I usedthem as the basis of a lecture before the Royal Institution onFebruary 9th, 1877. It appeared from these experiments that theoffspring did not tend to resemble their parent seeds in size, butto be always more mediocre than they—to be smaller than theparents, if the parents were large; to be larger than the parents, ifthe parents were very small…The experiments showed furtherthat the mean filial regression towards mediocrity was directlyproportional to the parental deviation from it.

Galton obviously expected his learned audience at the Royal Institution—the oldest independent research society in the world—to be as surprisedby his “noteworthy observation” as he had been. What is truly noteworthy isthat he was surprised by a statistical regularity that is as common as theair we breathe. Regression effects can be found wherever we look, but wedo not recognize them for what they are. They hide in plain sight. It tookGalton several years to work his way from his discovery of filial regressionin size to the broader notion that regression inevitably occurs when thecorrelation between two measures is less than perfect, and he needed thehelp of the most brilliant statisticians of his time to reach that conclusion.

One of the hurdles Galton had to overcome was the problem ofmeasuring regression between variables that are measured on differentscales, such as weight and piano playing. This is done by using thepopulation as a standard of reference. Imagine that weight and pianoplaying have been measured for 100 children in all grades of anelementary school, and that they have been ranked from high to low oneach measure. If Jane ranks third in piano playing and twenty-seventh inweight, it is appropriate to say that she is a better pianist than she is tall.

(PDF) Thinking, fast and slow by daniel kahneman - PDFSLIDE.US (177)

Let us make some assumptions that will simplify things:At any age,

Piano-playing success depends only on weekly hours of practice.Weight depends only on consumption of ice cream.Ice cream consumption and weekly hours of practice are unrelated.

Now, using ranks (or the standard scores that statisticians prefer), we canwrite some equations:

weight = age + ice cream consumptionpiano playing = age + weekly hours of practice

You can see that there will be regression to the mean when we predictpiano playing from weight, or vice versa. If all you know about Tom is thathe ranks twelfth in weight (well above average), you can infer (statistically)that he is probably older than average and also that he probably consumesmore ice cream than other children. If all you know about Barbara is thatshe is eighty-fifth in piano (far below the average of the group), you caninfer that she is likely to be young and that she is likely to practice less thanmost other children.

T h e correlation coefficient between two measures, which variesbetween 0 and 1, is a measure of the relative weight of the factors theyshare. For example, we all share half our genes with each of our parents,and for traits in which environmental factors have relatively little influence,such as height, the correlation between parent and child is not far from .50.To appreciate the meaning of the correlation measure, the following aresome examples of coefficients:

The correlation between the size of objects measured with precisionin English or in metric units is 1. Any factor that influences onemeasure also influences the other; 100% of determinants areshared.The correlation between self-reported height and weight among adultAmerican males is .41. If you included women and children, thecorrelation would be much higher, because individuals’ gender andage influence both their height ann wd their weight, boosting the

(PDF) Thinking, fast and slow by daniel kahneman - PDFSLIDE.US (178)

relative weight of shared factors.The correlation between SAT scores and college GPA isapproximately .60. However, the correlation between aptitude testsand success in graduate school is much lower, largely becausemeasured aptitude varies little in this selected group. If everyone hassimilar aptitude, differences in this measure are unlikely to play alarge role in measures of success.The correlation between income and education level in the UnitedStates is approximately .40.The correlation between family income and the last four digits of theirphone number is 0.

It took Francis Galton several years to figure out that correlation andregression are not two concepts—they are different perspectives on thesame concept. The general rule is straightforward but has surprisingconsequences: whenever the correlation between two scores is imperfect,there will be regression to the mean. To illustrate Galton’s insight, take aproposition that most people find quite interesting:

Highly intelligent women tend to marry men who are lessintelligent than they are.

You can get a good conversation started at a party by asking for anexplanation, and your friends will readily oblige. Even people who have hadsome exposure to statistics will spontaneously interpret the statement incausal terms. Some may think of highly intelligent women wanting to avoidthe competition of equally intelligent men, or being forced to compromisein their choice of spouse because intelligent men do not want to competewith intelligent women. More far-fetched explanations will come up at agood party. Now consider this statement:

The correlation between the intelligence scores of spouses isless than perfect.

This statement is obviously true and not interesting at all. Who wouldexpect the correlation to be perfect? There is nothing to explain. But thestatement you found interesting and the statement you found trivial arealgebraically equivalent. If the correlation between the intelligence ofspouses is less than perfect (and if men and women on average do notdiffer in intelligence), then it is a mathematical inevitability that highlyintelligent women will be married to husbands who are on average less

(PDF) Thinking, fast and slow by daniel kahneman - PDFSLIDE.US (179)

intelligent than they are (and vice versa, of course). The observedregression to the mean cannot be more interesting or more explainablethan the imperfect correlation.

You probably sympathize with Galton’s struggle with the concept ofregression. Indeed, the statistician David Freedman used to say that if thetopic of regression comes up in a criminal or civil trial, the side that mustexplain regression to the jury will lose the case. Why is it so hard? Themain reason for the difficulty is a recurrent theme of this book: our mind isstrongly biased toward causal explanations and does not deal well with“mere statistics.” When our attention is called to an event, associativememory will look for its cause—more precisely, activation will automaticallyspread to any cause that is already stored in memory. Causal explanationswill be evoked when regression is detected, but they will be wrongbecause the truth is that regression to the mean has an explanation butdoes not have a cause. The event that attracts our attention in the golfingtournament is the frequent deterioration of the performance of the golferswho werecte successful on day 1. The best explanation of it is that thosegolfers were unusually lucky that day, but this explanation lacks the causalforce that our minds prefer. Indeed, we pay people quite well to provideinteresting explanations of regression effects. A business commentatorwho correctly announces that “the business did better this year because ithad done poorly last year” is likely to have a short tenure on the air.

Our difficulties with the concept of regression originate with both System 1and System 2. Without special instruction, and in quite a few cases evenafter some statistical instruction, the relationship between correlation andregression remains obscure. System 2 finds it difficult to understand andlearn. This is due in part to the insistent demand for causal interpretations,which is a feature of System 1.

Depressed children treated with an energy drink improvesignificantly over a three-month period.

I made up this newspaper headline, but the fact it reports is true: if youtreated a group of depressed children for some time with an energy drink,they would show a clinically significant improvement. It is also the case thatdepressed children who spend some time standing on their head or hug acat for twenty minutes a day will also show improvement. Most readers ofsuch headlines will automatically infer that the energy drink or the cathugging caused an improvement, but this conclusion is completelyunjustified. Depressed children are an extreme group, they are more

(PDF) Thinking, fast and slow by daniel kahneman - PDFSLIDE.US (180)

depressed than most other children—and extreme groups regress to themean over time. The correlation between depression scores onsuccessive occasions of testing is less than perfect, so there will beregression to the mean: depressed children will get somewhat better overtime even if they hug no cats and drink no Red Bull. In order to concludethat an energy drink—or any other treatment—is effective, you mustcompare a group of patients who receive this treatment to a “control group”that receives no treatment (or, better, receives a placebo). The controlgroup is expected to improve by regression alone, and the aim of theexperiment is to determine whether the treated patients improve more thanregression can explain.

Incorrect causal interpretations of regression effects are not restricted toreaders of the popular press. The statistician Howard Wainer has drawnup a long list of eminent researchers who have made the same mistake—confusing mere correlation with causation. Regression effects are acommon source of trouble in research, and experienced scientists developa healthy fear of the trap of unwarranted causal inference.

One of my favorite examples of the errors of intuitive prediction is adaptedfrom Max Bazerman’s excellent text Judgment in Managerial DecisionMaking:

You are the sales forecaster for a department store chain. Allstores are similar in size and merchandise selection, but theirsales differ because of location, competition, and randomfactors. You are given the results for 2011 and asked to forecastsales for 2012. You have been instructed to accept the overallforecast of economists that sales will increase overall by 10%.How would you complete the following table?

Store 2011 20121 $11,000,000 ________2 $23,000,000 ________3 $18,000,000 ________4 $29,000,000 ________Total $61,000,000 $67,100,000

Having read this chapter, you know that the obvious solution of adding

(PDF) Thinking, fast and slow by daniel kahneman - PDFSLIDE.US (181)

10% to the sales of each store is wrong. You want your forecasts to beregressive, which requires adding more than 10% to the low-performingbranches and adding less (or even subtracting) to others. But if you askother people, you are likely to encounter puzzlement: Why do you botherthem with an obvious question? As Galton painfully discovered, theconcept of regression is far from obvious.

Speaking of Regression to Mediocrity

“She says experience has taught her that criticism is moreeffective than praise. What she doesn’t understand is that it’s alldue to regression to the mean.”

“Perhaps his second interview was less impressive than thefirst because he was afraid of disappointing us, but more likely itwas his first that was unusually good.”

“Our screening procedure is good but not perfect, so we shouldanticipate regression. We shouldn’t be surprised that the verybest candidates often fail to meet our expectations.”

(PDF) Thinking, fast and slow by daniel kahneman - PDFSLIDE.US (182)

Taming Intuitive Predictions

Life presents us with many occasions to forecast. Economists forecastinflation and unemployment, financial analysts forecast earnings, militaryexperts predict casualties, venture capitalists assess profitability,publishers and producers predict audiences, contractors estimate the timerequired to complete projects, chefs anticipate the demand for the disheson their menu, engineers estimate the amount of concrete needed for abuilding, fireground commanders assess the number of trucks that will beneeded to put out a fire. In our private lives, we forecast our spouse’sreaction to a proposed move or our own future adjustment to a new job.

Some predictive judgments, such as those made by engineers, relylargely on look-up tables, precise calculations, and explicit analyses ofoutcomes observed on similar occasions. Others involve intuition andSystem 1, in two main varieties. Some intuitions draw primarily on skill andexpertise acquired by repeated experience. The rapid and automaticjudgments and choices of chess masters, fireground commanders, andphysicians that Gary Klein has described in Sources of Power andelsewhere illustrate these skilled intuitions, in which a solution to the currentproblem comes to mind quickly because familiar cues are recognized.

Other intuitions, which are sometimes subjectively indistinguishable fromthe first, arise from the operation of heuristics that often substitute an easyquestion for the harder one that was asked. Intuitive judgments can bemade with high confidence even when they are based on nonregressiveassessments of weak evidence. Of course, many judgments, especially inthe professional domain, are influenced by a combination of analysis andintuition.

Nonregressive Intuitions

Let us return to a person we have already met:

Julie is currently a senior in a state university. She read fluentlywhen she was four years old. What is her grade point average(GPA)?

People who are familiar with the American educational scene quicklycome up with a number, which is often in the vicinity of 3.7 or 3.8. Howdoes this occur? Several operations of System 1 are involved.

(PDF) Thinking, fast and slow by daniel kahneman - PDFSLIDE.US (183)

A causal link between the evidence (Julie’s reading) and the target ofthe prediction (her GPA) is sought. The link can be indirect. In thisinstance, early reading and a high GDP are both indications ofacademic talent. Some connection is necessary. You (your System2) would probably reject as irrelevant a report of Julie winning a flyfishing competitiowhired D=n or excelling at weight lifting in highschool. The process is effectively dichotomous. We are capable ofrejecting information as irrelevant or false, but adjusting for smallerweaknesses in the evidence is not something that System 1 can do.As a result, intuitive predictions are almost completely insensitive tothe actual predictive quality of the evidence. When a link is found, asin the case of Julie’s early reading, WY SIATI applies: yourassociative memory quickly and automatically constructs the bestpossible story from the information available.Next, the evidence is evaluated in relation to a relevant norm. Howprecocious is a child who reads fluently at age four? What relativerank or percentile score corresponds to this achievement? Thegroup to which the child is compared (we call it a reference group) isnot fully specified, but this is also the rule in normal speech: ifsomeone graduating from college is described as “quite clever” yourarely need to ask, “When you say ‘quite clever,’ which referencegroup do you have in mind?”The next step involves substitution and intensity matching. Theevaluation of the flimsy evidence of cognitive ability in childhood issubstituted as an answer to the question about her college GPA.Julie will be assigned the same percentile score for her GPA and forher achievements as an early reader.The question specified that the answer must be on the GPA scale,which requires another intensity-matching operation, from a generalimpression of Julie’s academic achievements to the GPA thatmatches the evidence for her talent. The final step is a translation,from an impression of Julie’s relative academic standing to the GPAthat corresponds to it.

Intensity matching yields predictions that are as extreme as the evidenceon which they are based, leading people to give the same answer to twoquite different questions:

What is Julie’s percentile score on reading precocity?What is Julie’s percentile score on GPA?

(PDF) Thinking, fast and slow by daniel kahneman - PDFSLIDE.US (184)

By now you should easily recognize that all these operations arefeatures of System 1. I listed them here as an orderly sequence of steps,but of course the spread of activation in associative memory does notwork this way. You should imagine a process of spreading activation thatis initially prompted by the evidence and the question, feeds back uponitself, and eventually settles on the most coherent solution possible.

Amos and I once asked participants in an experiment to judgedescriptions of eight college freshmen, allegedly written by a counselor onthe basis of interviews of the entering class. Each description consisted offive adjectives, as in the following example:

intelligent, self-confident, well-read, hardworking, inquisitive

We asked some participants to answer two questions:

How much does this description impress you with respect toacademic ability?

What percentage of descriptions of freshmen do you believewould impress you more?

The questions require you to evaluate the evidence by comparing thedescription to your norm for descriptions of students by counselors. Thevery existence of such a norm is remarkable. Although you surely do notknow how you acquired it, you have a fairly clear sense of how muchenthusiasm the description conveys: the counselor believes that thisstudent is good, but not spectacularly good. There is room for strongeradjectives than intelligent (brilliant, creative), well-read (scholarly, erudite,impressively knowledgeable), and hardworking (passionate,perfectionist). The verdict: very likely to be in the top 15% but unlikely to bein the top 3%. There is impressive consensus in such judgments, at leastwithin a culture.

The other participants in our experiment were asked different questions:

What is your estimate of the grade point average that the studentwill obtain?What is the percentage of freshmen who obtain a higher GPA?

You need another look to detect the subtle difference between the two

(PDF) Thinking, fast and slow by daniel kahneman - PDFSLIDE.US (185)

sets of questions. The difference should be obvious, but it is not. Unlike thefirst questions, which required you only to evaluate the evidence, thesecond set involves a great deal of uncertainty. The question refers toactual performance at the end of the freshman year. What happenedduring the year since the interview was performed? How accurately canyou predict the student’s actual achievements in the first year at collegefrom five adjectives? Would the counselor herself be perfectly accurate ifshe predicted GPA from an interview?

The objective of this study was to compare the percentile judgments thatthe participants made when evaluating the evidence in one case, andwhen predicting the ultimate outcome in another. The results are easy tosummarize: the judgments were identical. Although the two sets ofquestions differ (one is about the description, the other about the student’sfuture academic performance), the participants treated them as if theywere the same. As was the case with Julie, the prediction of the future isnot distinguished from an evaluation of current evidence—predictionmatches evaluation. This is perhaps the best evidence we have for the roleof substitution. People are asked for a prediction but they substitute anevaluation of the evidence, without noticing that the question they answer isnot the one they were asked. This process is guaranteed to generatepredictions that are systematically biased; they completely ignoreregression to the mean.

During my military service in the Israeli Defense Forces, I spent sometime attached to a unit that selected candidates for officer training on thebasis of a series of interviews and field tests. The designated criterion forsuccessful prediction was a cadet’s final grade in officer school. Thevalidity of the ratings was known to be rather poor (I will tell more about it ina later chapter). The unit still existed years later, when I was a professorand collaborating with Amos in the study of intuitive judgment. I had goodcontacts with the people at the unit and asked them for a favor. In additionto the usual grading system they used to evaluate the candidates, I askedfor their best guess of the grade that each of the future cadets would obtainin officer school. They collected a few hundred such forecasts. The officerswho had produced the prediof рctions were all familiar with the lettergrading system that the school applied to its cadets and the approximateproportions of A’s, B’s, etc., among them. The results were striking: therelative frequency of A’s and B’s in the predictions was almost identical tothe frequencies in the final grades of the school.

These findings provide a compelling example of both substitution andintensity matching. The officers who provided the predictions completelyfailed to discriminate between two tasks:

(PDF) Thinking, fast and slow by daniel kahneman - PDFSLIDE.US (186)

their usual mission, which was to evaluate the performance ofcandidates during their stay at the unitthe task I had asked them to perform, which was an actual predictionof a future grade

They had simply translated their own grades onto the scale used in officerschool, applying intensity matching. Once again, the failure to address the(considerable) uncertainty of their predictions had led them to predictionsthat were completely nonregressive.

A Correction for Intuitive Predictions

Back to Julie, our precocious reader. The correct way to predict her GPAwas introduced in the preceding chapter. As I did there for golf onsuccessive days and for weight and piano playing, I write a schematicformula for the factors that determine reading age and college grades:

reading age = shared factors + factors specific to reading age =100%GPA = shared factors + factors specific to GPA = 100%

The shared factors involve genetically determined aptitude, the degree towhich the family supports academic interests, and anything else that wouldcause the same people to be precocious readers as children andacademically successful as young adults. Of course there are many factorsthat would affect one of these outcomes and not the other. Julie could havebeen pushed to read early by overly ambitious parents, she may have hadan unhappy love affair that depressed her college grades, she could havehad a skiing accident during adolescence that left her slightly impaired,and so on.

Recall that the correlation between two measures—in the present casereading age and GPA—is equal to the proportion of shared factors amongtheir determinants. What is your best guess about that proportion? Mymost optimistic guess is about 30%. Assuming this estimate, we have allwe need to produce an unbiased prediction. Here are the directions forhow to get there in four simple steps:

(PDF) Thinking, fast and slow by daniel kahneman - PDFSLIDE.US (187)

1. Start with an estimate of average GPA.2. Determine the GPA that matches your impression of the evidence.3. Estimate the correlation between your evidence and GPA.4. If the correlation is .30, move 30% of the distance from the average

to the matching GPA.

Step 1 gets you the baseline, the GPA you would have predicted if youwere told nothing about Julie beyond the fact that she is a graduatingsenior. In the absence of information, you would have predicted theaverage. (This is similar to assigning the base-rate probability of businessadministration grahavрduates when you are told nothing about Tom W.)Step 2 is your intuitive prediction, which matches your evaluation of theevidence. Step 3 moves you from the baseline toward your intuition, but thedistance you are allowed to move depends on your estimate of thecorrelation. You end up, at step 4, with a prediction that is influenced byyour intuition but is far more moderate.

This approach to prediction is general. You can apply it whenever youneed to predict a quantitative variable, such as GPA, profit from aninvestment, or the growth of a company. The approach builds on yourintuition, but it moderates it, regresses it toward the mean. When you havegood reasons to trust the accuracy of your intuitive prediction—a strongcorrelation between the evidence and the prediction—the adjustment willbe small.

Intuitive predictions need to be corrected because they are notregressive and therefore are biased. Suppose that I predict for each golferin a tournament that his score on day 2 will be the same as his score onday 1. This prediction does not allow for regression to the mean: thegolfers who fared well on day 1 will on average do less well on day 2, andthose who did poorly will mostly improve. When they are eventuallycompared to actual outcomes, nonregressive predictions will be found tobe biased. They are on average overly optimistic for those who did best onthe first day and overly pessimistic for those who had a bad start. Thepredictions are as extreme as the evidence. Similarly, if you use childhoodachievements to predict grades in college without regressing yourpredictions toward the mean, you will more often than not be disappointedby the academic outcomes of early readers and happily surprised by thegrades of those who learned to read relatively late. The corrected intuitivepredictions eliminate these biases, so that predictions (both high and low)are about equally likely to overestimate and to underestimate the truevalue. You still make errors when your predictions are unbiased, but theerrors are smaller and do not favor either high or low outcomes.

(PDF) Thinking, fast and slow by daniel kahneman - PDFSLIDE.US (188)

A Defense of Extreme Predictions?

I introduced Tom W earlier to illustrate predictions of discrete outcomessuch as field of specialization or success in an examination, which areexpressed by assigning a probability to a specified event (or in that caseby ranking outcomes from the most to the least probable). I also describeda procedure that counters the common biases of discrete prediction:neglect of base rates and insensitivity to the quality of information.

The biases we find in predictions that are expressed on a scale, such asGPA or the revenue of a firm, are similar to the biases observed in judgingthe probabilities of outcomes.

The corrective procedures are also similar:

Both contain a baseline prediction, which you would make if youknew nothing about the case at hand. In the categorical case, it wasthe base rate. In the numerical case, it is the average outcome in therelevant category.Both contain an intuitive prediction, which expresses the number thatcomes to your mind, whether it is a probability or a GPA.In both cases, you aim for a prediction that is intermediate betweenthe baseline and your intuitive response.In the default case of no useful evidence, you stay with the baseline.At the other extreme, you also stay with your initial predictiononsр.This will happen, of course, only if you remain completely confident inyour initial prediction after a critical review of the evidence thatsupports it.In most cases you will find some reason to doubt that the correlationbetween your intuitive judgment and the truth is perfect, and you willend up somewhere between the two poles.

This procedure is an approximation of the likely results of an appropriatestatistical analysis. If successful, it will move you toward unbiasedpredictions, reasonable assessments of probability, and moderatepredictions of numerical outcomes. The two procedures are intended toaddress the same bias: intuitive predictions tend to be overconfident andoverly extreme.

(PDF) Thinking, fast and slow by daniel kahneman - PDFSLIDE.US (189)

Correcting your intuitive predictions is a task for System 2. Significanteffort is required to find the relevant reference category, estimate thebaseline prediction, and evaluate the quality of the evidence. The effort isjustified only when the stakes are high and when you are particularly keennot to make mistakes. Furthermore, you should know that correcting yourintuitions may complicate your life. A characteristic of unbiased predictionsis that they permit the prediction of rare or extreme events only when theinformation is very good. If you expect your predictions to be of modestvalidity, you will never guess an outcome that is either rare or far from themean. If your predictions are unbiased, you will never have the satisfyingexperience of correctly calling an extreme case. You will never be able tosay, “I thought so!” when your best student in law school becomes aSupreme Court justice, or when a start-up that you thought very promisingeventually becomes a major commercial success. Given the limitations ofthe evidence, you will never predict that an outstanding high school studentwill be a straight-A student at Princeton. For the same reason, a venturecapitalist will never be told that the probability of success for a start-up inits early stages is “very high.”

The objections to the principle of moderating intuitive predictions mustbe taken seriously, because absence of bias is not always what mattersmost. A preference for unbiased predictions is justified if all errors ofprediction are treated alike, regardless of their direction. But there aresituations in which one type of error is much worse than another. When aventure capitalist looks for “the next big thing,” the risk of missing the nextGoogle or Facebook is far more important than the risk of making amodest investment in a start-up that ultimately fails. The goal of venturecapitalists is to call the extreme cases correctly, even at the cost ofoverestimating the prospects of many other ventures. For a conservativebanker making large loans, the risk of a single borrower going bankruptmay outweigh the risk of turning down several would-be clients who wouldfulfill their obligations. In such cases, the use of extreme language (“verygood prospect,” “serious risk of default”) may have some justification forthe comfort it provides, even if the information on which these judgmentsare based is of only modest validity.

For a rational person, predictions that are unbiased and moderateshould not present a problem. After all, the rational venture capitalist knowsthat even the most promising start-ups have only a moderate chance ofsuccess. She views her job as picking the most promising bets from thebets that are available and does not feel the need to delude herself aboutthe prospects of a start-up in which she plans to invest. Similarly, rationalindividuals predicting the revenue of a firm will not be bound to a singleys рnumber—they should consider the range of uncertainty around the most

(PDF) Thinking, fast and slow by daniel kahneman - PDFSLIDE.US (190)

likely outcome. A rational person will invest a large sum in an enterprisethat is most likely to fail if the rewards of success are large enough, withoutdeluding herself about the chances of success. However, we are not allrational, and some of us may need the security of distorted estimates toavoid paralysis. If you choose to delude yourself by accepting extremepredictions, however, you will do well to remain aware of your self-indulgence.

Perhaps the most valuable contribution of the corrective procedures Ipropose is that they will require you to think about how much you know. Iwill use an example that is familiar in the academic world, but theanalogies to other spheres of life are immediate. A department is about tohire a young professor and wants to choose the one whose prospects forscientific productivity are the best. The search committee has narroweddown the choice to two candidates:

Kim recently completed her graduate work. Herrecommendations are spectacular and she gave a brilliant talkand impressed everyone in her interviews. She has nosubstantial track record of scientific productivity.

Jane has held a postdoctoral position for the last three years.She has been very productive and her research record isexcellent, but her talk and interviews were less sparkling thanKim’s.

The intuitive choice favors Kim, because she left a stronger impression,and WYSIATI. But it is also the case that there is much less informationabout Kim than about Jane. We are back to the law of small numbers. Ineffect, you have a smaller sample of information from Kim than from Jane,and extreme outcomes are much more likely to be observed in smallsamples. There is more luck in the outcomes of small samples, and youshould therefore regress your prediction more deeply toward the mean inyour prediction of Kim’s future performance. When you allow for the factthat Kim is likely to regress more than Jane, you might end up selectingJane although you were less impressed by her. In the context of academicchoices, I would vote for Jane, but it would be a struggle to overcome myintuitive impression that Kim is more promising. Following our intuitions ismore natural, and somehow more pleasant, than acting against them.

You can readily imagine similar problems in different contexts, such as aventure capitalist choosing between investments in two start-ups thatoperate in different markets. One start-up has a product for which demand

(PDF) Thinking, fast and slow by daniel kahneman - PDFSLIDE.US (191)

can be estimated with fair precision. The other candidate is more excitingand intuitively promising, but its prospects are less certain. Whether thebest guess about the prospects of the second start-up is still superior whenthe uncertainty is factored in is a question that deserves carefulconsideration.

A Two-Systems View of Regression

Extreme predictions and a willingness to predict rare events from weakevidence are both manifestations of System 1. It is natural for theassociative machinery to match the extremeness of predictions to theperceived extremeness of evidence on which it is based—this is howsubstitution works. And it is natural for System 1 to generate overconfidentjudgments, because confidence, as we have seen, is determined by thecoherence of the best story you can tell from the evidence at hand. Bewarned: your intuitions will deliver predictions that are too extreme and youwill be inclinehe рd to put far too much faith in them.

Regression is also a problem for System 2. The very idea of regressionto the mean is alien and difficult to communicate and comprehend. Galtonhad a hard time before he understood it. Many statistics teachers dreadthe class in which the topic comes up, and their students often end up withonly a vague understanding of this crucial concept. This is a case whereSystem 2 requires special training. Matching predictions to the evidence isnot only something we do intuitively; it also seems a reasonable thing todo. We will not learn to understand regression from experience. Even whena regression is identified, as we saw in the story of the flight instructors, itwill be given a causal interpretation that is almost always wrong.

Speaking of Intuitive Predictions

“That start-up achieved an outstanding proof of concept, but weshouldn’t expect them to do as well in the future. They are still along way from the market and there is a lot of room forregression.”

“Our intuitive prediction is very favorable, but it is probably toohigh. Let’s take into account the strength of our evidence andregress the prediction toward the mean.”

(PDF) Thinking, fast and slow by daniel kahneman - PDFSLIDE.US (192)

“The investment may be a good idea, even if the best guess isthat it will fail. Let's not say we really believe it is the next Google.”

“I read one review of that brand and it was excellent. Still, thatcould have been a fluke. Let’s consider only the brands that havea large number of reviews and pick the one that looks best.”

(PDF) Thinking, fast and slow by daniel kahneman - PDFSLIDE.US (193)

Part 3

(PDF) Thinking, fast and slow by daniel kahneman - PDFSLIDE.US (194)

Overconfidence

(PDF) Thinking, fast and slow by daniel kahneman - PDFSLIDE.US (195)

The Illusion of Understanding

The trader-philosopher-statistician Nassim Taleb could also beconsidered a psychologist. In The Black Swan, Taleb introduced the notionof a narrative fallacy to describe how flawed stories of the past shape ourviews of the world and our expectations for the future. Narrative fallaciesarise inevitably from our continuous attempt to make sense of the world.The explanatory stories that people find compelling are simple; areconcrete rather than abstract; assign a larger role to talent, stupidity, andintentions than to luck; and focus on a few striking events that happenedrather than on the countless events that failed to happen. Any recent salientevent is a candidate to become the kernel of a causal narrative. Talebsuggests that we humans constantly fool ourselves by constructing flimsyaccounts of the past and believing they are true.

Good stories provide a simple and coherent account >A compelling narrative fosters an illusion of inevitability. Consider the

story of how Google turned into a giant of the technology industry. Twocreative graduate students in the computer science department atStanford University come up with a superior way of searching informationon the Internet. They seek and obtain funding to start a company and makea series of decisions that work out well. Within a few years, the companythey started is one of the most valuable stocks in America, and the twoformer graduate students are among the richest people on the planet. Onone memorable occasion, they were lucky, which makes the story evenmore compelling: a year after founding Google, they were willing to selltheir company for less than $1 million, but the buyer said the price was toohigh. Mentioning the single lucky incident actually makes it easier tounderestimate the multitude of ways in which luck affected the outcome.

A detailed history would specify the decisions of Google’s founders, butfor our purposes it suffices to say that almost every choice they made hada good outcome. A more complete narrative would describe the actions ofthe firms that Google defeated. The hapless competitors would appear tobe blind, slow, and altogether inadequate in dealing with the threat thateventually overwhelmed them.

I intentionally told this tale blandly, but you get the idea: there is a verygood story here. Fleshed out in more detail, the story could give you thesense that you understand what made Google succeed; it would alsomake you feel that you have learned a valuable general lesson about whatmakes businesses succeed. Unfortunately, there is good reason to believethat your sense of understanding and learning from the Google story islargely illusory. The ultimate test of an explanation is whether it would have

(PDF) Thinking, fast and slow by daniel kahneman - PDFSLIDE.US (196)

made the event predictable in advance. No story of Google’s unlikelysuccess will meet that test, because no story can include the myriad ofevents that would have caused a different outcome. The human mind doesnot deal well with nonevents. The fact that many of the important events thatdid occur involve choices further tempts you to exaggerate the role of skilland underestimate the part that luck played in the outcome. Because everycritical decision turned out well, the record suggests almost flawlessprescience—but bad luck could have disrupted any one of the successfulsteps. The halo effect adds the final touches, lending an aura of invincibilityto the heroes of the story.

Like watching a skilled rafter avoiding one potential calamity afteranother as he goes down the rapids, the unfolding of the Google story isthrilling because of the constant risk of disaster. However, there is foр aninstructive difference between the two cases. The skilled rafter has gonedown rapids hundreds of times. He has learned to read the roiling water infront of him and to anticipate obstacles. He has learned to make the tinyadjustments of posture that keep him upright. There are feweropportunities for young men to learn how to create a giant company, andfewer chances to avoid hidden rocks—such as a brilliant innovation by acompeting firm. Of course there was a great deal of skill in the Googlestory, but luck played a more important role in the actual event than it doesin the telling of it. And the more luck was involved, the less there is to belearned.

At work here is that powerful WY SIATI rule. You cannot help dealing withthe limited information you have as if it were all there is to know. You buildthe best possible story from the information available to you, and if it is agood story, you believe it. Paradoxically, it is easier to construct a coherentstory when you know little, when there are fewer pieces to fit into the puzzle.Our comforting conviction that the world makes sense rests on a securefoundation: our almost unlimited ability to ignore our ignorance.

I have heard of too many people who “knew well before it happened thatthe 2008 financial crisis was inevitable.” This sentence contains a highlyobjectionable word, which should be removed from our vocabulary indiscussions of major events. The word is, of course, knew. Some peoplethought well in advance that there would be a crisis, but they did not knowit. They now say they knew it because the crisis did in fact happen. This isa misuse of an important concept. In everyday language, we apply theword know only when what was known is true and can be shown to be true.We can know something only if it is both true and knowable. But the peoplewho thought there would be a crisis (and there are fewer of them than nowremember thinking it) could not conclusively show it at the time. Many

(PDF) Thinking, fast and slow by daniel kahneman - PDFSLIDE.US (197)

intelligent and well-informed people were keenly interested in the future ofthe economy and did not believe a catastrophe was imminent; I infer fromthis fact that the crisis was not knowable. What is perverse about the useof know in this context is not that some individuals get credit for presciencethat they do not deserve. It is that the language implies that the world ismore knowable than it is. It helps perpetuate a pernicious illusion.

The core of the illusion is that we believe we understand the past, whichimplies that the future also should be knowable, but in fact we understandthe past less than we believe we do. Know is not the only word that fostersthis illusion. In common usage, the words intuition and premonition alsoare reserved for past thoughts that turned out to be true. The statement “Ihad a premonition that the marriage would not last, but I was wrong”sounds odd, as does any sentence about an intuition that turned out to befalse. To think clearly about the future, we need to clean up the languagethat we use in labeling the beliefs we had in the past.

The Social Costs of Hindsight

The mind that makes up narratives about the past is a sense-makingorgan. When an unpredicted event occurs, we immediately adjust our viewof the world to accommodate the surprise. Imagine yourself before afootball game between two teams that have the same record of wins andlosses. Now the game is over, and one team trashed the other. In yourrevised model of the world, the winning team is much stronger than theloser, and your view of the past as well as of the future has been altered befрy that new perception. Learning from surprises is a reasonable thing todo, but it can have some dangerous consequences.

A general limitation of the human mind is its imperfect ability toreconstruct past states of knowledge, or beliefs that have changed. Onceyou adopt a new view of the world (or of any part of it), you immediatelylose much of your ability to recall what you used to believe before yourmind changed.

Many psychologists have studied what happens when people changetheir minds. Choosing a topic on which minds are not completely made up—say, the death penalty—the experimenter carefully measures people’sattitudes. Next, the participants see or hear a persuasive pro or conmessage. Then the experimenter measures people’s attitudes again; theyusually are closer to the persuasive message they were exposed to.Finally, the participants report the opinion they held beforehand. This taskturns out to be surprisingly difficult. Asked to reconstruct their formerbeliefs, people retrieve their current ones instead—an instance of

(PDF) Thinking, fast and slow by daniel kahneman - PDFSLIDE.US (198)

substitution—and many cannot believe that they ever felt differently.Your inability to reconstruct past beliefs will inevitably cause you to

underestimate the extent to which you were surprised by past events.Baruch Fischh off first demonstrated this “I-knew-it-all-along” effect, orhindsight bias, when he was a student in Jerusalem. Together with RuthBeyth (another of our students), Fischh off conducted a survey beforePresident Richard Nixon visited China and Russia in 1972. Therespondents assigned probabilities to fifteen possible outcomes ofNixon’s diplomatic initiatives. Would Mao Zedong agree to meet withNixon? Might the United States grant diplomatic recognition to China?After decades of enmity, could the United States and the Soviet Unionagree on anything significant?

After Nixon’s return from his travels, Fischh off and Beyth asked thesame people to recall the probability that they had originally assigned toeach of the fifteen possible outcomes. The results were clear. If an eventhad actually occurred, people exaggerated the probability that they hadassigned to it earlier. If the possible event had not come to pass, theparticipants erroneously recalled that they had always considered itunlikely. Further experiments showed that people were driven to overstatethe accuracy not only of their original predictions but also of those made byothers. Similar results have been found for other events that gripped publicattention, such as the O. J. Simpson murder trial and the impeachment ofPresident Bill Clinton. The tendency to revise the history of one’s beliefs inlight of what actually happened produces a robust cognitive illusion.

Hindsight bias has pernicious effects on the evaluations of decisionmakers. It leads observers to assess the quality of a decision not bywhether the process was sound but by whether its outcome was good orbad. Consider a low-risk surgical intervention in which an unpredictableaccident occurred that caused the patient’s death. The jury will be prone tobelieve, after the fact, that the operation was actually risky and that thedoctor who ordered it should have known better. This outcome bias makesit almost impossible to evaluate a decision properly—in terms of thebeliefs that were reasonable when the decision was made.

Hindsight is especially unkind to decision makers who act as agents forothers—physicians, financial advisers, third-base coaches, CEOs, socialworkers, diplomats, politicians. We are prone to blame decision makersfor good decisions that worked out badly and to give them too little creditfor successful movesecaр that appear obvious only after the fact. There isa clear outcome bias. When the outcomes are bad, the clients often blametheir agents for not seeing the handwriting on the wall—forgetting that itwas written in invisible ink that became legible only afterward. Actions that

(PDF) Thinking, fast and slow by daniel kahneman - PDFSLIDE.US (199)

seemed prudent in foresight can look irresponsibly negligent in hindsight.Based on an actual legal case, students in California were asked whetherthe city of Duluth, Minnesota, should have shouldered the considerablecost of hiring a full-time bridge monitor to protect against the risk thatdebris might get caught and block the free flow of water. One group wasshown only the evidence available at the time of the city’s decision; 24% ofthese people felt that Duluth should take on the expense of hiring a floodmonitor. The second group was informed that debris had blocked the river,causing major flood damage; 56% of these people said the city shouldhave hired the monitor, although they had been explicitly instructed not tolet hindsight distort their judgment.

The worse the consequence, the greater the hindsight bias. In the caseof a catastrophe, such as 9/11, we are especially ready to believe that theofficials who failed to anticipate it were negligent or blind. On July 10,2001, the Central Intelligence Agency obtained information that al-Qaedamight be planning a major attack against the United States. George Tenet,director of the CIA, brought the information not to President George W.Bush but to National Security Adviser Condoleezza Rice. When the factslater emerged, Ben Bradlee, the legendary executive editor of TheWashington Post, declared, “It seems to me elementary that if you’ve gotthe story that’s going to dominate history you might as well go right to thepresident.” But on July 10, no one knew—or could have known—that thistidbit of intelligence would turn out to dominate history.

Because adherence to standard operating procedures is difficult tosecond-guess, decision makers who expect to have their decisionsscrutinized with hindsight are driven to bureaucratic solutions—and to anextreme reluctance to take risks. As malpractice litigation became morecommon, physicians changed their procedures in multiple ways: orderedmore tests, referred more cases to specialists, applied conventionaltreatments even when they were unlikely to help. These actions protectedthe physicians more than they benefited the patients, creating the potentialfor conflicts of interest. Increased accountability is a mixed blessing.

Although hindsight and the outcome bias generally foster risk aversion,they also bring undeserved rewards to irresponsible risk seekers, such asa general or an entrepreneur who took a crazy gamble and won. Leaderswho have been lucky are never punished for having taken too much risk.Instead, they are believed to have had the flair and foresight to anticipatesuccess, and the sensible people who doubted them are seen in hindsightas mediocre, timid, and weak. A few lucky gambles can crown a recklessleader with a halo of prescience and boldness.

(PDF) Thinking, fast and slow by daniel kahneman - PDFSLIDE.US (200)

Recipes for Success

The sense-making machinery of System 1 makes us see the world asmore tidy, simple, predictable, and coherent than it really is. The illusionthat one has understood the past feeds the further illusion that one canpredict and control the future. These illusions are comforting. They reducethe anxiety that we would experience if we allowed ourselves to fullyacknowledge the uncertainties of existence. We all have a need for thereassuring message that actions have appropriate consequences, andthat success will reward wisdom and courage. Many bdecрusiness booksare tailor-made to satisfy this need.

Do leaders and management practices influence the outcomes of firmsin the market? Of course they do, and the effects have been confirmed bysystematic research that objectively assessed the characteristics of CEOsand their decisions, and related them to subsequent outcomes of the firm.In one study, the CEOs were characterized by the strategy of thecompanies they had led before their current appointment, as well as bymanagement rules and procedures adopted after their appointment. CEOsdo influence performance, but the effects are much smaller than a readingof the business press suggests.

Researchers measure the strength of relationships by a correlationcoefficient, which varies between 0 and 1. The coefficient was definedearlier (in relation to regression to the mean) by the extent to which twomeasures are determined by shared factors. A very generous estimate ofthe correlation between the success of the firm and the quality of its CEOmight be as high as .30, indicating 30% overlap. To appreciate thesignificance of this number, consider the following question:

Suppose you consider many pairs of firms. The two firms in eachpair are generally similar, but the CEO of one of them is betterthan the other. How often will you find that the firm with thestronger CEO is the more successful of the two?

In a well-ordered and predictable world, the correlation would be perfect(1), and the stronger CEO would be found to lead the more successful firmin 100% of the pairs. If the relative success of similar firms was determinedentirely by factors that the CEO does not control (call them luck, if youwish), you would find the more successful firm led by the weaker CEO 50%of the time. A correlation of .30 implies that you would find the strongerCEO leading the stronger firm in about 60% of the pairs—an improvementof a mere 10 percentage points over random guessing, hardly grist for the

(PDF) Thinking, fast and slow by daniel kahneman - PDFSLIDE.US (201)

hero worship of CEOs we so often witness.If you expected this value to be higher—and most of us do—then you

should take that as an indication that you are prone to overestimate thepredictability of the world you live in. Make no mistake: improving the oddsof success from 1:1 to 3:2 is a very significant advantage, both at theracetrack and in business. From the perspective of most business writers,however, a CEO who has so little control over performance would not beparticularly impressive even if her firm did well. It is difficult to imaginepeople lining up at airport bookstores to buy a book that enthusiasticallydescribes the practices of business leaders who, on average, dosomewhat better than chance. Consumers have a hunger for a clearmessage about the determinants of success and failure in business, andthey need stories that offer a sense of understanding, however illusory.

In his penetrating book The Halo Effect, Philip Rosenzweig, a businessschool professor based in Switzerland, shows how the demand for illusorycertainty is met in two popular genres of business writing: histories of therise (usually) and fall (occasionally) of particular individuals andcompanies, and analyses of differences between successful and lesssuccessful firms. He concludes that stories of success and failureconsistently exaggerate the impact of leadership style and managementpractices on firm outcomes, and thus their message is rarely useful.

To appreciate what is going on, imagine that business experts, such asother CEOs, are asked to comment on the reputation of the chief executiveof a company. They poрare keenly aware of whether the company hasrecently been thriving or failing. As we saw earlier in the case of Google,this knowledge generates a halo. The CEO of a successful company islikely to be called flexible, methodical, and decisive. Imagine that a yearhas passed and things have gone sour. The same executive is nowdescribed as confused, rigid, and authoritarian. Both descriptions soundright at the time: it seems almost absurd to call a successful leader rigidand confused, or a struggling leader flexible and methodical.

Indeed, the halo effect is so powerful that you probably find yourselfresisting the idea that the same person and the same behaviors appearmethodical when things are going well and rigid when things are goingpoorly. Because of the halo effect, we get the causal relationshipbackward: we are prone to believe that the firm fails because its CEO isrigid, when the truth is that the CEO appears to be rigid because the firm isfailing. This is how illusions of understanding are born.

The halo effect and outcome bias combine to explain the extraordinaryappeal of books that seek to draw operational morals from systematicexamination of successful businesses. One of the best-known examples of

(PDF) Thinking, fast and slow by daniel kahneman - PDFSLIDE.US (202)

this genre is Jim Collins and Jerry I. Porras’s Built to Last. The bookcontains a thorough analysis of eighteen pairs of competing companies, inwhich one was more successful than the other. The data for thesecomparisons are ratings of various aspects of corporate culture, strategy,and management practices. “We believe every CEO, manager, andentrepreneur in the world should read this book,” the authors proclaim.“You can build a visionary company.”

The basic message of Built to Last and other similar books is that goodmanagerial practices can be identified and that good practices will berewarded by good results. Both messages are overstated. Thecomparison of firms that have been more or less successful is to asignificant extent a comparison between firms that have been more or lesslucky. Knowing the importance of luck, you should be particularlysuspicious when highly consistent patterns emerge from the comparison ofsuccessful and less successful firms. In the presence of randomness,regular patterns can only be mirages.

Because luck plays a large role, the quality of leadership andmanagement practices cannot be inferred reliably from observations ofsuccess. And even if you had perfect foreknowledge that a CEO hasbrilliant vision and extraordinary competence, you still would be unable topredict how the company will perform with much better accuracy than theflip of a coin. On average, the gap in corporate profitability and stockreturns between the outstanding firms and the less successful firms studiedin Built to Last shrank to almost nothing in the period following the study.The average profitability of the companies identified in the famous InSearch of Excellence dropped sharply as well within a short time. A studyo f Fortune’s “Most Admired Companies” finds that over a twenty-yearperiod, the firms with the worst ratings went on to earn much higher stockreturns than the most admired firms.

You are probably tempted to think of causal explanations for theseobservations: perhaps the successful firms became complacent, the lesssuccessful firms tried harder. But this is the wrong way to think about whathappened. The average gap must shrink, because the original gap wasdue in good part to luck, which contributed both to the success of the topfirms and to the lagging performance of the rest. We have alreadyencountered this statistical fact of life: regression to the mean.

Stories of how businesses rise and fall strike a chord with readers byoffering what the human mind needs: a simple message of triumph andfailure that identifies clear causes and ignores the determinative power ofluck and the inevitability of regression. These stories induce and maintainan illusion of understanding, imparting lessons of little enduring value to

(PDF) Thinking, fast and slow by daniel kahneman - PDFSLIDE.US (203)

readers who are all too eager to believe them.

Speaking of Hindsight

“The mistake appears obvious, but it is just hindsight. You couldnot have known in advance.”

“He’s learning too much from this success story, which is too tidy.He has fallen for a narrative fallacy.”

“She has no evidence for saying that the firm is badly managed.All she knows is that its stock has gone down. This is an outcomebias, part hindsight and part halo effect.”

“Let’s not fall for the outcome bias. This was a stupid decisioneven though it worked out well.”

(PDF) Thinking, fast and slow by daniel kahneman - PDFSLIDE.US (204)

The Illusion of Validity

System 1 is designed to jump to conclusions from little evidence—and it isnot designed to know the size of its jumps. Because of WYSIATI, only theevidence at hand counts. Because of confidence by coherence, thesubjective confidence we have in our opinions reflects the coherence of thestory that System 1 and System 2 have constructed. The amount ofevidence and its quality do not count for much, because poor evidence canmake a very good story. For some of our most important beliefs we haveno evidence at all, except that people we love and trust hold these beliefs.Considering how little we know, the confidence we have in our beliefs ispreposterous—and it is also essential.

The Illusion of Validity

Many decades ago I spent what seemed like a great deal of time under ascorching sun, watching groups of sweaty soldiers as they solved aproblem. I was doing my national service in the Israeli Army at the time. Ihad completed an undergraduate degree in psychology, and after a yearas an infantry officer was assigned to the army’s Psychology Branch,where one of my occasional duties was to help evaluate candidates forofficer training. We used methods that had been developed by the BritishArmy in World War II.

One test, called the “leaderless group challenge,” was conducted on anobstacle field. Eight candidates, strangers to each other, with all insignia ofrank removed and only numbered tags to identify them, were instructed tolift a long log from the ground and haul it to a wall about six feet high. Theentire group had to get to the other side of the wall without the log touchingeither the ground or the wall, and without anyone touching the wall. If any ofthese things happened, they had to declare itsigрЉ T and start again.

There was more than one way to solve the problem. A common solutionwas for the team to send several men to the other side by crawling over thepole as it was held at an angle, like a giant fishing rod, by other membersof the group. Or else some soldiers would climb onto someone’s shouldersand jump across. The last man would then have to jump up at the pole, heldup at an angle by the rest of the group, shinny his way along its length asthe others kept him and the pole suspended in the air, and leap safely tothe other side. Failure was common at this point, which required them tostart all over again.

As a colleague and I monitored the exercise, we made note of who tookcharge, who tried to lead but was rebuffed, how cooperative each soldier

(PDF) Thinking, fast and slow by daniel kahneman - PDFSLIDE.US (205)

was in contributing to the group effort. We saw who seemed to bestubborn, submissive, arrogant, patient, hot-tempered, persistent, or aquitter. We sometimes saw competitive spite when someone whose ideahad been rejected by the group no longer worked very hard. And we sawreactions to crisis: who berated a comrade whose mistake had caused thewhole group to fail, who stepped forward to lead when the exhausted teamhad to start over. Under the stress of the event, we felt, each man’s truenature revealed itself. Our impression of each candidate’s character wasas direct and compelling as the color of the sky.

After watching the candidates make several attempts, we had tosummarize our impressions of soldiers’ leadership abilities anddetermine, with a numerical score, who should be eligible for officertraining. We spent some time discussing each case and reviewing ourimpressions. The task was not difficult, because we felt we had alreadyseen each soldier’s leadership skills. Some of the men had looked likestrong leaders, others had seemed like wimps or arrogant fools, othersmediocre but not hopeless. Quite a few looked so weak that we ruled themout as candidates for officer rank. When our multiple observations of eachcandidate converged on a coherent story, we were completely confident inour evaluations and felt that what we had seen pointed directly to the future.The soldier who took over when the group was in trouble and led the teamover the wall was a leader at that moment. The obvious best guess abouthow he would do in training, or in combat, was that he would be aseffective then as he had been at the wall. Any other prediction seemedinconsistent with the evidence before our eyes.

Because our impressions of how well each soldier had performed weregenerally coherent and clear, our formal predictions were just as definite. Asingle score usually came to mind and we rarely experienced doubts orformed conflicting impressions. We were quite willing to declare, “This onewill never make it,” “That fellow is mediocre, but he should do okay,” or “Hewill be a star.” We felt no need to question our forecasts, moderate them,or equivocate. If challenged, however, we were prepared to admit, “But ofcourse anything could happen.” We were willing to make that admissionbecause, despite our definite impressions about individual candidates, weknew with certainty that our forecasts were largely useless.

The evidence that we could not forecast success accurately wasoverwhelming. Every few months we had a feedback session in which welearned how the cadets were doing at the officer-training school and couldcompare our assessments against the opinions of commanders who hadbeen monitoring them for some time. The story was always the same: ourability to predict performance at the school was negligible. Our forecastswere better than blind guesses, but not by much.

(PDF) Thinking, fast and slow by daniel kahneman - PDFSLIDE.US (206)

We weed re downcast for a while after receiving the discouragingnews. But this was the army. Useful or not, there was a routine to befollowed and orders to be obeyed. Another batch of candidates arrived thenext day. We took them to the obstacle field, we faced them with the wall,they lifted the log, and within a few minutes we saw their true naturesrevealed, as clearly as before. The dismal truth about the quality of ourpredictions had no effect whatsoever on how we evaluated candidates andvery little effect on the confidence we felt in our judgments and predictionsabout individuals.

What happened was remarkable. The global evidence of our previousfailure should have shaken our confidence in our judgments of thecandidates, but it did not. It should also have caused us to moderate ourpredictions, but it did not. We knew as a general fact that our predictionswere little better than random guesses, but we continued to feel and act asif each of our specific predictions was valid. I was reminded of the Müller-Lyer illusion, in which we know the lines are of equal length yet still seethem as being different. I was so struck by the analogy that I coined a termfor our experience: the illusion of validity.

I had discovered my first cognitive illusion.

Decades later, I can see many of the central themes of my thinking—and ofthis book—in that old story. Our expectations for the soldiers’ futureperformance were a clear instance of substitution, and of therepresentativeness heuristic in particular. Having observed one hour of asoldier’s behavior in an artificial situation, we felt we knew how well hewould face the challenges of officer training and of leadership in combat.Our predictions were completely nonregressive—we had no reservationsabout predicting failure or outstanding success from weak evidence. Thiswas a clear instance of WYSIATI. We had compelling impressions of thebehavior we observed and no good way to represent our ignorance of thefactors that would eventually determine how well the candidate wouldperform as an officer.

Looking back, the most striking part of the story is that our knowledge ofthe general rule—that we could not predict—had no effect on ourconfidence in individual cases. I can see now that our reaction was similarto that of Nisbett and Borgida’s students when they were told that mostpeople did not help a stranger suffering a seizure. They certainly believedthe statistics they were shown, but the base rates did not influence theirjudgment of whether an individual they saw on the video would or would nothelp a stranger. Just as Nisbett and Borgida showed, people are often

(PDF) Thinking, fast and slow by daniel kahneman - PDFSLIDE.US (207)

reluctant to infer the particular from the general.Subjective confidence in a judgment is not a reasoned evaluation of the

probability that this judgment is correct. Confidence is a feeling, whichreflects the coherence of the information and the cognitive ease ofprocessing it. It is wise to take admissions of uncertainty seriously, butdeclarations of high confidence mainly tell you that an individual hasconstructed a coherent story in his mind, not necessarily that the story istrue.

The Illusion of Stock-Picking Skill

In 1984, Amos and I and our friend Richard Thaler visited a Wall Streetfirm. Our host, a senior investment manager, had invited us to discuss therole of judgment biases in investing. I knew so little about finance that I didnot even know what to ask him, but I remember one exchange. “When yousell a stock,” d n I asked, “who buys it?” He answered with a wave in thevague direction of the window, indicating that he expected the buyer to besomeone else very much like him. That was odd: What made one personbuy and the other sell? What did the sellers think they knew that the buyersdid not?

Since then, my questions about the stock market have hardened into alarger puzzle: a major industry appears to be built largely on an illusion ofskill. Billions of shares are traded every day, with many people buyingeach stock and others selling it to them. It is not unusual for more than 100million shares of a single stock to change hands in one day. Most of thebuyers and sellers know that they have the same information; theyexchange the stocks primarily because they have different opinions. Thebuyers think the price is too low and likely to rise, while the sellers think theprice is high and likely to drop. The puzzle is why buyers and sellers alikethink that the current price is wrong. What makes them believe they knowmore about what the price should be than the market does? For most ofthem, that belief is an illusion.

In its broad outlines, the standard theory of how the stock market worksis accepted by all the participants in the industry. Everybody in theinvestment business has read Burton Malkiel’s wonderful book A RandomWalk Down Wall Street. Malkiel’s central idea is that a stock’s priceincorporates all the available knowledge about the value of the companyand the best predictions about the future of the stock. If some peoplebelieve that the price of a stock will be higher tomorrow, they will buy moreof it today. This, in turn, will cause its price to rise. If all assets in a marketare correctly priced, no one can expect either to gain or to lose by trading.

(PDF) Thinking, fast and slow by daniel kahneman - PDFSLIDE.US (208)

Perfect prices leave no scope for cleverness, but they also protect foolsfrom their own folly. We now know, however, that the theory is not quiteright. Many individual investors lose consistently by trading, anachievement that a dart-throwing chimp could not match. The firstdemonstration of this startling conclusion was collected by Terry Odean, afinance professor at UC Berkeley who was once my student.

Odean began by studying the trading records of 10,000 brokerageaccounts of individual investors spanning a seven-year period. He wasable to analyze every transaction the investors executed through that firm,nearly 163,000 trades. This rich set of data allowed Odean to identify allinstances in which an investor sold some of his holdings in one stock andsoon afterward bought another stock. By these actions the investorrevealed that he (most of the investors were men) had a definite ideaabout the future of the two stocks: he expected the stock that he chose tobuy to do better than the stock he chose to sell.

To determine whether those ideas were well founded, Odean comparedthe returns of the stock the investor had sold and the stock he had boughtin its place, over the course of one year after the transaction. The resultswere unequivocally bad. On average, the shares that individual traderssold did better than those they bought, by a very substantial margin: 3.2percentage points per year, above and beyond the significant costs ofexecuting the two trades.

It is important to remember that this is a statement about averages:some individuals did much better, others did much worse. However, it isclear that for the large majority of individual investors, taking a shower anddoing nothing would have been a better policy than implementing the ideasthat came to their minds. Later research by Odean and his colleague BradBarber supported this conclusion. In a paper titled “Trading Is Hazardousto Yourt-t Wealth,” they showed that, on average, the most active tradershad the poorest results, while the investors who traded the least earned thehighest returns. In another paper, titled “Boys Will Be Boys,” they showedthat men acted on their useless ideas significantly more often than women,and that as a result women achieved better investment results than men.

Of course, there is always someone on the other side of eachtransaction; in general, these are financial institutions and professionalinvestors, who are ready to take advantage of the mistakes that individualtraders make in choosing a stock to sell and another stock to buy. Furtherresearch by Barber and Odean has shed light on these mistakes.Individual investors like to lock in their gains by selling “winners,” stocksthat have appreciated since they were purchased, and they hang on totheir losers. Unfortunately for them, recent winners tend to do better thanrecent losers in the short run, so individuals sell the wrong stocks. They

(PDF) Thinking, fast and slow by daniel kahneman - PDFSLIDE.US (209)

also buy the wrong stocks. Individual investors predictably flock tocompanies that draw their attention because they are in the news.Professional investors are more selective in responding to news. Thesefindings provide some justification for the label of “smart money” thatfinance professionals apply to themselves.

Although professionals are able to extract a considerable amount ofwealth from amateurs, few stock pickers, if any, have the skill needed tobeat the market consistently, year after year. Professional investors,including fund managers, fail a basic test of skill: persistent achievement.The diagnostic for the existence of any skill is the consistency of individualdifferences in achievement. The logic is simple: if individual differences inany one year are due entirely to luck, the ranking of investors and funds willvary erratically and the year-to-year correlation will be zero. Where there isskill, however, the rankings will be more stable. The persistence ofindividual differences is the measure by which we confirm the existence ofskill among golfers, car salespeople, orthodontists, or speedy tollcollectors on the turnpike.

Mutual funds are run by highly experienced and hardworkingprofessionals who buy and sell stocks to achieve the best possible resultsfor their clients. Nevertheless, the evidence from more than fifty years ofresearch is conclusive: for a large majority of fund managers, the selectionof stocks is more like rolling dice than like playing poker. Typically at leasttwo out of every three mutual funds underperform the overall market in anygiven year.

More important, the year-to-year correlation between the outcomes ofmutual funds is very small, barely higher than zero. The successful funds inany given year are mostly lucky; they have a good roll of the dice. There isgeneral agreement among researchers that nearly all stock pickers,whether they know it or not—and few of them do—are playing a game ofchance. The subjective experience of traders is that they are makingsensible educated guesses in a situation of great uncertainty. In highlyefficient markets, however, educated guesses are no more accurate thanblind guesses.

Some years ago I had an unusual opportunity to examine the illusion offinancial skill up close. I had been invited to speak to a group of investmentadvisers in a firm that provided financial advice and other services to verywealthy clients. I asked for some data to prepare my presentation and wasgranted a small treasure: a spreadsheet summarizing the investmentoutcomes of some twenty-five anonymous wealth advisers, for each of

(PDF) Thinking, fast and slow by daniel kahneman - PDFSLIDE.US (210)

eight consecutive years. Each adviser’s scoof re for each year was his(most of them were men) main determinant of his year-end bonus. It was asimple matter to rank the advisers by their performance in each year andto determine whether there were persistent differences in skill among themand whether the same advisers consistently achieved better returns fortheir clients year after year.

To answer the question, I computed correlation coefficients between therankings in each pair of years: year 1 with year 2, year 1 with year 3, andso on up through year 7 with year 8. That yielded 28 correlationcoefficients, one for each pair of years. I knew the theory and wasprepared to find weak evidence of persistence of skill. Still, I was surprisedto find that the average of the 28 correlations was .01. In other words, zero.The consistent correlations that would indicate differences in skill were notto be found. The results resembled what you would expect from a dice-rolling contest, not a game of skill.

No one in the firm seemed to be aware of the nature of the game that itsstock pickers were playing. The advisers themselves felt they werecompetent professionals doing a serious job, and their superiors agreed.On the evening before the seminar, Richard Thaler and I had dinner withsome of the top executives of the firm, the people who decide on the sizeof bonuses. We asked them to guess the year-to-year correlation in therankings of individual advisers. They thought they knew what was comingand smiled as they said “not very high” or “performance certainlyfluctuates.” It quickly became clear, however, that no one expected theaverage correlation to be zero.

Our message to the executives was that, at least when it came tobuilding portfolios, the firm was rewarding luck as if it were skill. Thisshould have been shocking news to them, but it was not. There was nosign that they disbelieved us. How could they? After all, we had analyzedtheir own results, and they were sophisticated enough to see theimplications, which we politely refrained from spelling out. We all went oncalmly with our dinner, and I have no doubt that both our findings and theirimplications were quickly swept under the rug and that life in the firm wenton just as before. The illusion of skill is not only an individual aberration; itis deeply ingrained in the culture of the industry. Facts that challenge suchbasic assumptions—and thereby threaten people’s livelihood and self-esteem—are simply not absorbed. The mind does not digest them. This isparticularly true of statistical studies of performance, which provide base-rate information that people generally ignore when it clashes with theirpersonal impressions from experience.

The next morning, we reported the findings to the advisers, and theirresponse was equally bland. Their own experience of exercising careful

(PDF) Thinking, fast and slow by daniel kahneman - PDFSLIDE.US (211)

judgment on complex problems was far more compelling to them than anobscure statistical fact. When we were done, one of the executives I haddined with the previous evening drove me to the airport. He told me, with atrace of defensiveness, “I have done very well for the firm and no one cantake that away from me.” I smiled and said nothing. But I thought, “Well, Itook it away from you this morning. If your success was due mostly tochance, how much credit are you entitled to take for it?”

What Supports the Illusions of Skill and Validity?

Cognitive illusions can be more stubborn than visual illusions. What youlearned about the Müller-Lyer illusion did not change the way you see thelines, but it changed your behavior. You now know that you cannot trust yourimpression of the lenglli th of lines that have fins appended to them, andyou also know that in the standard Müller-Lyer display you cannot trust whatyou see. When asked about the length of the lines, you will report yourinformed belief, not the illusion that you continue to see. In contrast, whenmy colleagues and I in the army learned that our leadership assessmenttests had low validity, we accepted that fact intellectually, but it had noimpact on either our feelings or our subsequent actions. The response weencountered in the financial firm was even more extreme. I am convincedthat the message that Thaler and I delivered to both the executives and theportfolio managers was instantly put away in a dark corner of memorywhere it would cause no damage.

Why do investors, both amateur and professional, stubbornly believe thatthey can do better than the market, contrary to an economic theory thatmost of them accept, and contrary to what they could learn from adispassionate evaluation of their personal experience? Many of thethemes of previous chapters come up again in the explanation of theprevalence and persistence of an illusion of skill in the financial world.

The most potent psychological cause of the illusion is certainly that thepeople who pick stocks are exercising high-level skills. They consulteconomic data and forecasts, they examine income statements andbalance sheets, they evaluate the quality of top management, and theyassess the competition. All this is serious work that requires extensivetraining, and the people who do it have the immediate (and valid)experience of using these skills. Unfortunately, skill in evaluating thebusiness prospects of a firm is not sufficient for successful stock trading,where the key question is whether the information about the firm is alreadyincorporated in the price of its stock. Traders apparently lack the skill toanswer this crucial question, but they appear to be ignorant of their

(PDF) Thinking, fast and slow by daniel kahneman - PDFSLIDE.US (212)

ignorance. As I had discovered from watching cadets on the obstacle field,subjective confidence of traders is a feeling, not a judgment. Ourunderstanding of cognitive ease and associative coherence locatessubjective confidence firmly in System 1.

Finally, the illusions of validity and skill are supported by a powerfulprofessional culture. We know that people can maintain an unshakablefaith in any proposition, however absurd, when they are sustained by acommunity of like-minded believers. Given the professional culture of thefinancial community, it is not surprising that large numbers of individuals inthat world believe themselves to be among the chosen few who can dowhat they believe others cannot.

The Illusions of Pundits

The idea that the future is unpredictable is undermined every day by theease with which the past is explained. As Nassim Taleb pointed out in TheBlack Swan, our tendency to construct and believe coherent narratives ofthe past makes it difficult for us to accept the limits of our forecastingability. Everything makes sense in hindsight, a fact that financial punditsexploit every evening as they offer convincing accounts of the day’s events.And we cannot suppress the powerful intuition that what makes sense inhindsight today was predictable yesterday. The illusion that we understandthe past fosters overconfidence in our ability to predict the future.

The often-used image of the “march of history” implies order anddirection. Marches, unlike strolls or walks, are not random. We think thatwe should be able to explain the past by focusing on either large socialmovements and cultural and technological developments or the intentionsand abilities of a few g co reat men. The idea that large historical eventsare determined by luck is profoundly shocking, although it is demonstrablytrue. It is hard to think of the history of the twentieth century, including itslarge social movements, without bringing in the role of Hitler, Stalin, andMao Zedong. But there was a moment in time, just before an egg wasfertilized, when there was a fifty-fifty chance that the embryo that becameHitler could have been a female. Compounding the three events, there wasa probability of one-eighth of a twentieth century without any of the threegreat villains and it is impossible to argue that history would have beenroughly the same in their absence. The fertilization of these three eggs hadmomentous consequences, and it makes a joke of the idea that long-termdevelopments are predictable.

Yet the illusion of valid prediction remains intact, a fact that is exploitedby people whose business is prediction—not only financial experts but

(PDF) Thinking, fast and slow by daniel kahneman - PDFSLIDE.US (213)

pundits in business and politics, too. Television and radio stations andnewspapers have their panels of experts whose job it is to comment on therecent past and foretell the future. Viewers and readers have theimpression that they are receiving information that is somehow privileged,or at least extremely insightful. And there is no doubt that the pundits andtheir promoters genuinely believe they are offering such information. PhilipTetlock, a psychologist at the University of Pennsylvania, explained theseso-called expert predictions in a landmark twenty-year study, which hepublished in his 2005 book Expert Political Judgment: How Good Is It?How Can We Know? Tetlock has set the terms for any future discussion ofthis topic.

Tetlock interviewed 284 people who made their living “commenting oroffering advice on political and economic trends.” He asked them toassess the probabilities that certain events would occur in the not toodistant future, both in areas of the world in which they specialized and inregions about which they had less knowledge. Would Gorbachev beousted in a coup? Would the United States go to war in the Persian Gulf?Which country would become the next big emerging market? In all, Tetlockgathered more than 80,000 predictions. He also asked the experts howthey reached their conclusions, how they reacted when proved wrong, andhow they evaluated evidence that did not support their positions.Respondents were asked to rate the probabilities of three alternativeoutcomes in every case: the persistence of the status quo, more ofsomething such as political freedom or economic growth, or less of thatthing.

The results were devastating. The experts performed worse than theywould have if they had simply assigned equal probabilities to each of thethree potential outcomes. In other words, people who spend their time, andearn their living, studying a particular topic produce poorer predictions thandart-throwing monkeys who would have distributed their choices evenlyover the options. Even in the region they knew best, experts were notsignificantly better than nonspecialists.

Those who know more forecast very slightly better than those who knowless. But those with the most knowledge are often less reliable. The reasonis that the person who acquires more knowledge develops an enhancedillusion of her skill and becomes unrealistically overconfident. “We reachthe point of diminishing marginal predictive returns for knowledgedisconcertingly quickly,” Tetlock writes. “In this age of academichyperspecialization, there is no reason for supposing that contributors totop journals—distinguished political scientists, area study specialists,economists, and so on—are any better than journalists or attentive readers

(PDF) Thinking, fast and slow by daniel kahneman - PDFSLIDE.US (214)

o f The New York Times in ‘reading&#oul 8217; emerging situations.”The more famous the forecaster, Tetlock discovered, the more flamboyantthe forecasts. “Experts in demand,” he writes, “were more overconfidentthan their colleagues who eked out existences far from the limelight.”

Tetlock also found that experts resisted admitting that they had beenwrong, and when they were compelled to admit error, they had a largecollection of excuses: they had been wrong only in their timing, anunforeseeable event had intervened, or they had been wrong but for theright reasons. Experts are just human in the end. They are dazzled by theirown brilliance and hate to be wrong. Experts are led astray not by whatthey believe, but by how they think, says Tetlock. He uses the terminologyfrom Isaiah Berlin’s essay on Tolstoy, “The Hedgehog and the Fox.”Hedgehogs “know one big thing” and have a theory about the world; theyaccount for particular events within a coherent framework, bristle withimpatience toward those who don’t see things their way, and are confidentin their forecasts. They are also especially reluctant to admit error. Forhedgehogs, a failed prediction is almost always “off only on timing” or “verynearly right.” They are opinionated and clear, which is exactly whattelevision producers love to see on programs. Two hedgehogs on differentsides of an issue, each attacking the idiotic ideas of the adversary, makefor a good show.

Foxes, by contrast, are complex thinkers. They don’t believe that one bigthing drives the march of history (for example, they are unlikely to acceptthe view that Ronald Reagan single-handedly ended the cold war bystanding tall against the Soviet Union). Instead the foxes recognize thatreality emerges from the interactions of many different agents and forces,including blind luck, often producing large and unpredictable outcomes. Itwas the foxes who scored best in Tetlock’s study, although theirperformance was still very poor. They are less likely than hedgehogs to beinvited to participate in television debates.

It is Not the Experts’ Fault—The World is Difficult

The main point of this chapter is not that people who attempt to predict thefuture make many errors; that goes without saying. The first lesson is thaterrors of prediction are inevitable because the world is unpredictable. Thesecond is that high subjective confidence is not to be trusted as anindicator of accuracy (low confidence could be more informative).

Short-term trends can be forecast, and behavior and achievements canbe predicted with fair accuracy from previous behaviors andachievements. But we should not expect performance in officer training

(PDF) Thinking, fast and slow by daniel kahneman - PDFSLIDE.US (215)

and in combat to be predictable from behavior on an obstacle field—behavior both on the test and in the real world is determined by manyfactors that are specific to the particular situation. Remove one highlyassertive member from a group of eight candidates and everyone else’spersonalities will appear to change. Let a sniper’s bullet move by a fewcentimeters and the performance of an officer will be transformed. I do notdeny the validity of all tests—if a test predicts an important outcome with avalidity of .20 or .30, the test should be used. But you should not expectmore. You should expect little or nothing from Wall Street stock pickerswho hope to be more accurate than the market in predicting the future ofprices. And you should not expect much from pundits making long-termforecasts—although they may have valuable insights into the near future.The line that separates the possibly predictable future from theunpredictable distant future is in yet to be drawn.

Speaking of Illusory Skill

“He knows that the record indicates that the development of thisillness is mostly unpredictable. How can he be so confident in thiscase? Sounds like an illusion of validity.”

“She has a coherent story that explains all she knows, and thecoherence makes her feel good.”

“What makes him believe that he is smarter than the market? Isthis an illusion of skill?”

“She is a hedgehog. She has a theory that explains everything,and it gives her the illusion that she understands the world.”

“The question is not whether these experts are well trained. It iswhether their world is predictable.”

(PDF) Thinking, fast and slow by daniel kahneman - PDFSLIDE.US (216)

Intuitions vs. Formulas

Paul Meehl was a strange and wonderful character, and one of the mostversatile psychologists of the twentieth century. Among the departments inwhich he had faculty appointments at the University of Minnesota werepsychology, law, psychiatry, neurology, and philosophy. He also wrote onreligion, political science, and learning in rats. A statistically sophisticatedresearcher and a fierce critic of empty claims in clinical psychology, Meehlwas also a practicing psychoanalyst. He wrote thoughtful essays on thephilosophical foundations of psychological research that I almostmemorized while I was a graduate student. I never met Meehl, but he wasone of my heroes from the time I read his Clinical vs. StatisticalPrediction: A Theoretical Analysis and a Review of the Evidence.

In the slim volume that he later called “my disturbing little book,” Meehlreviewed the results of 20 studies that had analyzed whether clinicalpredictions based on the subjective impressions of trained professionalswere more accurate than statistical predictions made by combining a fewscores or ratings according to a rule. In a typical study, trained counselorspredicted the grades of freshmen at the end of the school year. Thecounselors interviewed each student for forty-five minutes. They also hadaccess to high school grades, several aptitude tests, and a four-pagepersonal statement. The statistical algorithm used only a fraction of thisinformation: high school grades and one aptitude test. Nevertheless, theformula was more accurate than 11 of the 14 counselors. Meehl reportedgenerally similar results across a variety of other forecast outcomes,including violations of parole, success in pilot training, and criminalrecidivism.

Not surprisingly, Meehl’s book provoked shock and disbelief amongclinical psychologists, and the controversy it started has engendered astream of research that is still flowing today, more than fifty yephy Љdiars after its publication. The number of studies reporting comparisons ofclinical and statistical predictions has increased to roughly two hundred,but the score in the contest between algorithms and humans has notchanged. About 60% of the studies have shown significantly betteraccuracy for the algorithms. The other comparisons scored a draw inaccuracy, but a tie is tantamount to a win for the statistical rules, which arenormally much less expensive to use than expert judgment. No exceptionhas been convincingly documented.

The range of predicted outcomes has expanded to cover medicalvariables such as the longevity of cancer patients, the length of hospitalstays, the diagnosis of cardiac disease, and the susceptibility of babies to

(PDF) Thinking, fast and slow by daniel kahneman - PDFSLIDE.US (217)

sudden infant death syndrome; economic measures such as the prospectsof success for new businesses, the evaluation of credit risks by banks, andthe future career satisfaction of workers; questions of interest togovernment agencies, including assessments of the suitability of fosterparents, the odds of recidivism among juvenile offenders, and thelikelihood of other forms of violent behavior; and miscellaneous outcomessuch as the evaluation of scientific presentations, the winners of footballgames, and the future prices of Bordeaux wine. Each of these domainsentails a significant degree of uncertainty and unpredictability. Wedescribe them as “low-validity environments.” In every case, the accuracyof experts was matched or exceeded by a simple algorithm.

As Meehl pointed out with justified pride thirty years after the publicationof his book, “There is no controversy in social science which shows such alarge body of qualitatively diverse studies coming out so uniformly in thesame direction as this one.”

The Princeton economist and wine lover Orley Ashenfelter has offered acompelling demonstration of the power of simple statistics to outdo world-renowned experts. Ashenfelter wanted to predict the future value of fineBordeaux wines from information available in the year they are made. Thequestion is important because fine wines take years to reach their peakquality, and the prices of mature wines from the same vineyard varydramatically across different vintages; bottles filled only twelve monthsapart can differ in value by a factor of 10 or more. An ability to forecastfuture prices is of substantial value, because investors buy wine, like art, inthe anticipation that its value will appreciate.

It is generally agreed that the effect of vintage can be due only tovariations in the weather during the grape-growing season. The best winesare produced when the summer is warm and dry, which makes theBordeaux wine industry a likely beneficiary of global warming. The industryis also helped by wet springs, which increase quantity without much effecton quality. Ashenfelter converted that conventional knowledge into astatistical formula that predicts the price of a wine—for a particularproperty and at a particular age—by three features of the weather: theaverage temperature over the summer growing season, the amount of rainat harvest-time, and the total rainfall during the previous winter. His formulaprovides accurate price forecasts years and even decades into the future.Indeed, his formula forecasts future prices much more accurately than thecurrent prices of young wines do. This new example of a “Meehl pattern”challenges the abilities of the experts whose opinions help shape the earlyprice. It also challenges economic theory, according to which prices shouldreflect all the available information, including the weather. Ashenfelter’sformula is extremely accurate—the correlation between his predictions and

(PDF) Thinking, fast and slow by daniel kahneman - PDFSLIDE.US (218)

actual prices is above .90.Why are experts e yinferior to algorithms? One reason, which Meehl

suspected, is that experts try to be clever, think outside the box, andconsider complex combinations of features in making their predictions.Complexity may work in the odd case, but more often than not it reducesvalidity. Simple combinations of features are better. Several studies haveshown that human decision makers are inferior to a prediction formulaeven when they are given the score suggested by the formula! They feelthat they can overrule the formula because they have additional informationabout the case, but they are wrong more often than not. According toMeehl, there are few circumstances under which it is a good idea tosubstitute judgment for a formula. In a famous thought experiment, hedescribed a formula that predicts whether a particular person will go to themovies tonight and noted that it is proper to disregard the formula ifinformation is received that the individual broke a leg today. The name“broken-leg rule” has stuck. The point, of course, is that broken legs arevery rare—as well as decisive.

Another reason for the inferiority of expert judgment is that humans areincorrigibly inconsistent in making summary judgments of complexinformation. When asked to evaluate the same information twice, theyfrequently give different answers. The extent of the inconsistency is often amatter of real concern. Experienced radiologists who evaluate chest X-rays as “normal” or “abnormal” contradict themselves 20% of the timewhen they see the same picture on separate occasions. A study of 101independent auditors who were asked to evaluate the reliability of internalcorporate audits revealed a similar degree of inconsistency. A review of41 separate studies of the reliability of judgments made by auditors,pathologists, psychologists, organizational managers, and otherprofessionals suggests that this level of inconsistency is typical, even whena case is reevaluated within a few minutes. Unreliable judgments cannotbe valid predictors of anything.

The widespread inconsistency is probably due to the extreme contextdependency of System 1. We know from studies of priming that unnoticedstimuli in our environment have a substantial influence on our thoughts andactions. These influences fluctuate from moment to moment. The briefpleasure of a cool breeze on a hot day may make you slightly morepositive and optimistic about whatever you are evaluating at the time. Theprospects of a convict being granted parole may change significantlyduring the time that elapses between successive food breaks in the parolejudges’ schedule. Because you have little direct knowledge of what goeson in your mind, you will never know that you might have made a different

(PDF) Thinking, fast and slow by daniel kahneman - PDFSLIDE.US (219)

judgment or reached a different decision under very slightly differentcircumstances. Formulas do not suffer from such problems. Given thesame input, they always return the same answer. When predictability ispoor—which it is in most of the studies reviewed by Meehl and hisfollowers—inconsistency is destructive of any predictive validity.

The research suggests a surprising conclusion: to maximize predictiveaccuracy, final decisions should be left to formulas, especially in low-validity environments. In admission decisions for medical schools, forexample, the final determination is often made by the faculty members whointerview the candidate. The evidence is fragmentary, but there are solidgrounds for a conjecture: conducting an interview is likely to diminish theaccuracy of a selection procedure, if the interviewers also make the finaladmission decisions. Because interviewers are overconfident in theirintuitions, they will assign too much weight to their personal impressionsand too little weight to other sources of information, lowering validity.Similarly, the experts who evaluate the quas plity of immature wine topredict its future have a source of information that almost certainly makesthings worse rather than better: they can taste the wine. In addition, ofcourse, even if they have a good understanding of the effects of theweather on wine quality, they will not be able to maintain the consistency ofa formula.

The most important development in the field since Meehl’s original work isRobyn Dawes’s famous article “The Robust Beauty of Improper LinearModels in Decision Making.” The dominant statistical practice in the socialsciences is to assign weights to the different predictors by following analgorithm, called multiple regression, that is now built into conventionalsoftware. The logic of multiple regression is unassailable: it finds theoptimal formula for putting together a weighted combination of thepredictors. However, Dawes observed that the complex statisticalalgorithm adds little or no value. One can do just as well by selecting a setof scores that have some validity for predicting the outcome and adjustingthe values to make them comparable (by using standard scores or ranks).A formula that combines these predictors with equal weights is likely to bejust as accurate in predicting new cases as the multiple-regression formulathat was optimal in the original sample. More recent research went further:formulas that assign equal weights to all the predictors are often superior,because they are not affected by accidents of sampling.

The surprising success of equal-weighting schemes has an importantpractical implication: it is possible to develop useful algorithms without anyprior statistical research. Simple equally weighted formulas based on

(PDF) Thinking, fast and slow by daniel kahneman - PDFSLIDE.US (220)

existing statistics or on common sense are often very good predictors ofsignificant outcomes. In a memorable example, Dawes showed thatmarital stability is well predicted by a formula:

frequency of lovemaking minus frequency of quarrels

You don’t want your result to be a negative number.The important conclusion from this research is that an algorithm that is

constructed on the back of an envelope is often good enough to competewith an optimally weighted formula, and certainly good enough to outdoexpert judgment. This logic can be applied in many domains, ranging fromthe selection of stocks by portfolio managers to the choices of medicaltreatments by doctors or patients.

A classic application of this approach is a simple algorithm that hassaved the lives of hundreds of thousands of infants. Obstetricians hadalways known that an infant who is not breathing normally within a fewminutes of birth is at high risk of brain damage or death. Until theanesthesiologist Virginia Apgar intervened in 1953, physicians andmidwives used their clinical judgment to determine whether a baby was indistress. Different practitioners focused on different cues. Some watchedfor breathing problems while others monitored how soon the baby cried.Without a standardized procedure, danger signs were often missed, andmany newborn infants died.

One day over breakfast, a medical resident asked how Dr. Apgar wouldmake a systematic assessment of a newborn. “That’s easy,” she replied.“You would do it like this.” Apgar jotted down five variables (heart rate,respiration, reflex, muscle tone, and color) and three scores (0, 1, or 2,depending on the robustness of each sign). Realizing that she might havemade a breakequthrough that any delivery room could implement, Apgarbegan rating infants by this rule one minute after they were born. A babywith a total score of 8 or above was likely to be pink, squirming, crying,grimacing, with a pulse of 100 or more—in good shape. A baby with ascore of 4 or below was probably bluish, flaccid, passive, with a slow orweak pulse—in need of immediate intervention. Applying Apgar’s score,the staff in delivery rooms finally had consistent standards for determiningwhich babies were in trouble, and the formula is credited for an importantcontribution to reducing infant mortality. The Apgar test is still used everyday in every delivery room. Atul Gawande’s recent A Checklist Manifestoprovides many other examples of the virtues of checklists and simple rules.

(PDF) Thinking, fast and slow by daniel kahneman - PDFSLIDE.US (221)

The Hostility to Algorithms

From the very outset, clinical psychologists responded to Meehl’s ideaswith hostility and disbelief. Clearly, they were in the grip of an illusion of skillin terms of their ability to make long-term predictions. On reflection, it iseasy to see how the illusion came about and easy to sympathize with theclinicians’ rejection of Meehl’s research.

The statistical evidence of clinical inferiority contradicts clinicians’everyday experience of the quality of their judgments. Psychologists whowork with patients have many hunches during each therapy session,anticipating how the patient will respond to an intervention, guessing whatwill happen next. Many of these hunches are confirmed, illustrating thereality of clinical skill.

The problem is that the correct judgments involve short-term predictionsin the context of the therapeutic interview, a skill in which therapists mayhave years of practice. The tasks at which they fail typically require long-term predictions about the patient’s future. These are much more difficult,even the best formulas do only modestly well, and they are also tasks thatthe clinicians have never had the opportunity to learn properly—they wouldhave to wait years for feedback, instead of receiving the instantaneousfeedback of the clinical session. However, the line between what clinicianscan do well and what they cannot do at all well is not obvious, and certainlynot obvious to them. They know they are skilled, but they don’t necessarilyknow the boundaries of their skill. Not surprisingly, then, the idea that amechanical combination of a few variables could outperform the subtlecomplexity of human judgment strikes experienced clinicians as obviouslywrong.

The debate about the virtues of clinical and statistical prediction hasalways had a moral dimension. The statistical method, Meehl wrote, wascriticized by experienced clinicians as “mechanical, atomistic, additive, cutand dried, artificial, unreal, arbitrary, incomplete, dead, pedantic,fractionated, trivial, forced, static, superficial, rigid, sterile, academic,pseudoscientific and blind.” The clinical method, on the other hand, waslauded by its proponents as “dynamic, global, meaningful, holistic, subtle,sympathetic, configural, patterned, organized, rich, deep, genuine,sensitive, sophisticated, real, living, concrete, natural, true to life, andunderstanding.”

This is an attitude we can all recognize. When a human competes with amachine, whether it is John Henry a-hammerin’ on the mountain or thechess genius Garry Kasparov facing off against the computer Deep Blue,our sympathies lie with our fellow human. The aversion to algorithms

(PDF) Thinking, fast and slow by daniel kahneman - PDFSLIDE.US (222)

making decisions that affect humans is rooted in the strong preference thatmany people have for the ormnatural over the synthetic or artificial. Askedwhether they would rather eat an organic or a commercially grown apple,most people prefer the “all natural” one. Even after being informed that thetwo apples taste the same, have identical nutritional value, and are equallyhealthful, a majority still prefer the organic fruit. Even the producers of beerhave found that they can increase sales by putting “All Natural” or “NoPreservatives” on the label.

The deep resistance to the demystification of expertise is illustrated bythe reaction of the European wine community to Ashenfelter’s formula forpredicting the price of Bordeaux wines. Ashenfelter’s formula answered aprayer: one might thus have expected that wine lovers everywhere wouldbe grateful to him for demonstrably improving their ability to identify thewines that later would be good. Not so. The response in French winecircles, wrote The New York Times, ranged “somewhere between violentand hysterical.” Ashenfelter reports that one oenophile called his findings“ludicrous and absurd.” Another scoffed, “It is like judging movies withoutactually seeing them.”

The prejudice against algorithms is magnified when the decisions areconsequential. Meehl remarked, “I do not quite know how to alleviate thehorror some clinicians seem to experience when they envisage a treatablecase being denied treatment because a ‘blind, mechanical’ equationmisclassifies him.” In contrast, Meehl and other proponents of algorithmshave argued strongly that it is unethical to rely on intuitive judgments forimportant decisions if an algorithm is available that will make fewermistakes. Their rational argument is compelling, but it runs against astubborn psychological reality: for most people, the cause of a mistakematters. The story of a child dying because an algorithm made a mistakeis more poignant than the story of the same tragedy occurring as a result ofhuman error, and the difference in emotional intensity is readily translatedinto a moral preference.

Fortunately, the hostility to algorithms will probably soften as their role ineveryday life continues to expand. Looking for books or music we mightenjoy, we appreciate recommendations generated by soft ware. We take itfor granted that decisions about credit limits are made without the directintervention of any human judgment. We are increasingly exposed toguidelines that have the form of simple algorithms, such as the ratio ofgood and bad cholesterol levels we should strive to attain. The public isnow well aware that formulas may do better than humans in some criticaldecisions in the world of sports: how much a professional team should payfor particular rookie players, or when to punt on fourth down. Theexpanding list of tasks that are assigned to algorithms should eventually

(PDF) Thinking, fast and slow by daniel kahneman - PDFSLIDE.US (223)

expanding list of tasks that are assigned to algorithms should eventuallyreduce the discomfort that most people feel when they first encounter thepattern of results that Meehl described in his disturbing little book.

Learning from Meehl

In 1955, as a twenty-one-year-old lieutenant in the Israeli Defense Forces, Iwas assigned to set up an interview system for the entire army. If youwonder why such a responsibility would be forced upon someone soyoung, bear in mind that the state of Israel itself was only seven years old atthe time; all its institutions were under construction, and someone had tobuild them. Odd as it sounds today, my bachelor’s degree in psychologyprobably qualified me as the best-trained psychologist in the army. Mydirect supervisor, a brilliant researcher, had a degree in chemistry.

An idilnterview routine was already in place when I was given mymission. Every soldier drafted into the army completed a battery ofpsychometric tests, and each man considered for combat duty wasinterviewed for an assessment of personality. The goal was to assign therecruit a score of general fitness for combat and to find the best match ofhis personality among various branches: infantry, artillery, armor, and soon. The interviewers were themselves young draftees, selected for thisassignment by virtue of their high intelligence and interest in dealing withpeople. Most were women, who were at the time exempt from combatduty. Trained for a few weeks in how to conduct a fifteen- to twenty-minuteinterview, they were encouraged to cover a range of topics and to form ageneral impression of how well the recruit would do in the army.

Unfortunately, follow-up evaluations had already indicated that thisinterview procedure was almost useless for predicting the future successof recruits. I was instructed to design an interview that would be moreuseful but would not take more time. I was also told to try out the newinterview and to evaluate its accuracy. From the perspective of a seriousprofessional, I was no more qualified for the task than I was to build abridge across the Amazon.

Fortunately, I had read Paul Meehl’s “little book,” which had appearedjust a year earlier. I was convinced by his argument that simple, statisticalrules are superior to intuitive “clinical” judgments. I concluded that the thencurrent interview had failed at least in part because it allowed theinterviewers to do what they found most interesting, which was to learnabout the dynamics of the interviewee’s mental life. Instead, we should usethe limited time at our disposal to obtain as much specific information aspossible about the interviewee’s life in his normal environment. Anotherlesson I learned from Meehl was that we should abandon the procedure in

(PDF) Thinking, fast and slow by daniel kahneman - PDFSLIDE.US (224)

which the interviewers’ global evaluations of the recruit determined the finaldecision. Meehl’s book suggested that such evaluations should not betrusted and that statistical summaries of separately evaluated attributeswould achieve higher validity.

I decided on a procedure in which the interviewers would evaluateseveral relevant personality traits and score each separately. The finalscore of fitness for combat duty would be computed according to astandard formula, with no further input from the interviewers. I made up alist of six characteristics that appeared relevant to performance in acombat unit, including “responsibility,” “sociability,” and “masculine pride.” Ithen composed, for each trait, a series of factual questions about theindividual’s life before his enlistment, including the number of different jobshe had held, how regular and punctual he had been in his work or studies,the frequency of his interactions with friends, and his interest andparticipation in sports, among others. The idea was to evaluate asobjectively as possible how well the recruit had done on each dimension.

By focusing on standardized, factual questions, I hoped to combat thehalo effect, where favorable first impressions influence later judgments. Asa further precaution against halos, I instructed the interviewers to gothrough the six traits in a fixed sequence, rating each trait on a five-pointscale before going on to the next. And that was that. I informed theinterviewers that they need not concern themselves with the recruit’s futureadjustment to the military. Their only task was to elicit relevant facts abouthis past and to use that information to score each personality dimension.“Your function is to provide reliable measurements,” I told them. “Leave thepredicok tive validity to me,” by which I meant the formula that I was goingto devise to combine their specific ratings.

The interviewers came close to mutiny. These bright young people weredispleased to be ordered, by someone hardly older than themselves, toswitch off their intuition and focus entirely on boring factual questions. Oneof them complained, “You are turning us into robots!” So I compromised.“Carry out the interview exactly as instructed,” I told them, “and when youare done, have your wish: close your eyes, try to imagine the recruit as asoldier, and assign him a score on a scale of 1 to 5.”

Several hundred interviews were conducted by this new method, and afew months later we collected evaluations of the soldiers’ performancefrom the commanding officers of the units to which they had beenassigned. The results made us happy. As Meehl’s book had suggested,the new interview procedure was a substantial improvement over the oldone. The sum of our six ratings predicted soldiers’ performance muchmore accurately than the global evaluations of the previous interviewingmethod, although far from perfectly. We had progressed from “completely

(PDF) Thinking, fast and slow by daniel kahneman - PDFSLIDE.US (225)

useless” to “moderately useful.”The big surprise to me was that the intuitive judgment that the

interviewers summoned up in the “close your eyes” exercise also did verywell, indeed just as well as the sum of the six specific ratings. I learnedfrom this finding a lesson that I have never forgotten: intuition adds valueeven in the justly derided selection interview, but only after a disciplinedcollection of objective information and disciplined scoring of separatetraits. I set a formula that gave the “close your eyes” evaluation the sameweight as the sum of the six trait ratings. A more general lesson that Ilearned from this episode was do not simply trust intuitive judgment—yourown or that of others—but do not dismiss it, either.

Some forty-five years later, after I won a Nobel Prize in economics, I wasfor a short time a minor celebrity in Israel. On one of my visits, someonehad the idea of escorting me around my old army base, which still housedthe unit that interviews new recruits. I was introduced to the commandingofficer of the Psychological Unit, and she described their currentinterviewing practices, which had not changed much from the system I haddesigned; there was, it turned out, a considerable amount of researchindicating that the interviews still worked well. As she came to the end ofher description of how the interviews are conducted, the officer added,“And then we tell them, ‘Close your eyes.’”

Do It Yourself

The message of this chapter is readily applicable to tasks other thanmaking manpower decisions for an army. Implementing interviewprocedures in the spirit of Meehl and Dawes requires relatively little effortbut substantial discipline. Suppose that you need to hire a salesrepresentative for your firm. If you are serious about hiring the bestpossible person for the job, this is what you should do. First, select a fewtraits that are prerequisites for success in this position (technicalproficiency, engaging personality, reliability, and so on). Don’t overdo it—six dimensions is a good number. The traits you choose should be asindependent as possible from each other, and you should feel that you canassess them reliably by asking a few factual questions. Next, make a list ofthose questions for each trait and think about how you will score it, say ona 1–5 scale. You should have an idea of what you will caleigl “very weak” or“very strong.”

These preparations should take you half an hour or so, a smallinvestment that can make a significant difference in the quality of thepeople you hire. To avoid halo effects, you must collect the information on

(PDF) Thinking, fast and slow by daniel kahneman - PDFSLIDE.US (226)

one trait at a time, scoring each before you move on to the next one. Donot skip around. To evaluate each candidate, add up the six scores.Because you are in charge of the final decision, you should not do a “closeyour eyes.” Firmly resolve that you will hire the candidate whose final scoreis the highest, even if there is another one whom you like better—try toresist your wish to invent broken legs to change the ranking. A vast amountof research offers a promise: you are much more likely to find the bestcandidate if you use this procedure than if you do what people normally doin such situations, which is to go into the interview unprepared and to makechoices by an overall intuitive judgment such as “I looked into his eyes andliked what I saw.”

Speaking of Judges vs. Formulas

“Whenever we can replace human judgment by a formula, weshould at least consider it.”

“He thinks his judgments are complex and subtle, but a simplecombination of scores could probably do better.”

“Let’s decide in advance what weight to give to the data we haveon the candidates’ past performance. Otherwise we will give toomuch weight to our impression from the interviews.”

(PDF) Thinking, fast and slow by daniel kahneman - PDFSLIDE.US (227)

Expert Intuition: When Can We Trust It?

Professional controversies bring out the worst in academics. Scientificjournals occasionally publish exchanges, often beginning with someone’scritique of another’s research, followed by a reply and a rejoinder. I havealways thought that these exchanges are a waste of time. Especially whenthe original critique is sharply worded, the reply and the rejoinder are oftenexercises in what I have called sarcasm for beginners and advancedsarcasm. The replies rarely concede anything to a biting critique, and it isalmost unheard of for a rejoinder to admit that the original critique wasmisguided or erroneous in any way. On a few occasions I have respondedto criticisms that I thought were grossly misleading, because a failure torespond can be interpreted as conceding error, but I have never found thehostile exchanges instructive. In search of another way to deal withdisagreements, I have engaged in a few “adversarial collaborations,” inwhich scholars who disagree on the science agree to write a jointlyauthored paper on their differences, and sometimes conduct researchtogether. In especially tense situations, the research is moderated by anarbiter.

My most satisfying and productive adversarial collaboration was withGary Klein, the intellectual leader of an association of scholars andpractitioners who do not like the kind of work I do. They call themselvesstudents of Naturalistic Decision Making, or NDM, and mostly work inorganizations where the"0%Љ ty often study how experts work. The NDMers adamantly reject the focus on biases in the heuristics and biasesapproach. They criticize this model as overly concerned with failures anddriven by artificial experiments rather than by the study of real people doingthings that matter. They are deeply skeptical about the value of using rigidalgorithms to replace human judgment, and Paul Meehl is not among theirheroes. Gary Klein has eloquently articulated this position over manyyears.

This is hardly the basis for a beautiful friendship, but there is more to thestory. I had never believed that intuition is always misguided. I had alsobeen a fan of Klein’s studies of expertise in firefighters since I first saw adraft of a paper he wrote in the 1970s, and was impressed by his bookSources of Power, much of which analyzes how experienced professionalsdevelop intuitive skills. I invited him to join in an effort to map the boundarythat separates the marvels of intuition from its flaws. He was intrigued bythe idea and we went ahead with the project—with no certainty that it wouldsucceed. We set out to answer a specific question: When can you trust anexperienced professional who claims to have an intuition? It was obvious

(PDF) Thinking, fast and slow by daniel kahneman - PDFSLIDE.US (228)

that Klein would be more disposed to be trusting, and I would be moreskeptical. But could we agree on principles for answering the generalquestion?

Over seven or eight years we had many discussions, resolved manydisagreements, almost blew up more than once, wrote many draft s,became friends, and eventually published a joint article with a title that tellsthe story: “Conditions for Intuitive Expertise: A Failure to Disagree.”Indeed, we did not encounter real issues on which we disagreed—but wedid not really agree.

Marvels and Flaws

Malcolm Gladwell’s bestseller Blink appeared while Klein and I wereworking on the project, and it was reassuring to find ourselves inagreement about it. Gladwell’s book opens with the memorable story of artexperts faced with an object that is described as a magnificent example ofa kouros, a sculpture of a striding boy. Several of the experts had strongvisceral reactions: they felt in their gut that the statue was a fake but werenot able to articulate what it was about it that made them uneasy. Everyonewho read the book—millions did—remembers that story as a triumph ofintuition. The experts agreed that they knew the sculpture was a fakewithout knowing how they knew—the very definition of intuition. The storyappears to imply that a systematic search for the cue that guided theexperts would have failed, but Klein and I both rejected that conclusion.From our point of view, such an inquiry was needed, and if it had beenconducted properly (which Klein knows how to do), it would probably havesucceeded.

Although many readers of the kouros example were surely drawn to analmost magical view of expert intuition, Gladwell himself does not hold thatposition. In a later chapter he describes a massive failure of intuition:Americans elected President Harding, whose only qualification for theposition was that he perfectly looked the part. Square jawed and tall, hewas the perfect image of a strong and decisive leader. People voted forsomeone who looked strong and decisive without any other reason tobelieve that he was. An intuitive prediction of how Harding would performas president arose from substituting one question for another. A reader ofthis book should expect such an intuition to be held with confidence.

Intuition as Recognition

The early experiences that shaped Klein’s views of intuition were starkly

(PDF) Thinking, fast and slow by daniel kahneman - PDFSLIDE.US (229)

different from mine. My thinking was formed by observing the illusion ofvalidity in myself and by reading Paul Meehl’s demonstrations of theinferiority of clinical prediction. In contrast, Klein’s views were shaped byhis early studies of fireground commanders (the leaders of firefightingteams). He followed them as they fought fires and later interviewed theleader about his thoughts as he made decisions. As Klein described it inour joint article, he and his collaborators

investigated how the commanders could make good decisionswithout comparing options. The initial hypothesis was thatcommanders would restrict their analysis to only a pair of options,but that hypothesis proved to be incorrect. In fact, thecommanders usually generated only a single option, and that wasall they needed. They could draw on the repertoire of patterns thatthey had compiled during more than a decade of both real andvirtual experience to identify a plausible option, which theyconsidered first. They evaluated this option by mentally simulatingit to see if it would work in the situation they were facing…. If thecourse of action they were considering seemed appropriate, theywould implement it. If it had shortcomings, they would modify it. Ifthey could not easily modify it, they would turn to the next mostplausible option and run through the same procedure until anacceptable course of action was found.

Klein elaborated this description into a theory of decision making that hecalled the recognition-primed decision (RPD) model, which applies tofirefighters but also describes expertise in other domains, including chess.The process involves both System 1 and System 2. In the first phase, atentative plan comes to mind by an automatic function of associativememory—System 1. The next phase is a deliberate process in which theplan is mentally simulated to check if it will work—an operation of System2. The model of intuitive decision making as pattern recognition developsideas presented some time ago by Herbert Simon, perhaps the onlyscholar who is recognized and admired as a hero and founding figure byall the competing clans and tribes in the study of decision making. I quotedHerbert Simon’s definition of intuition in the introduction, but it will makemore sense when I repeat it now: “The situation has provided a cue; thiscue has given the expert access to information stored in memory, and theinformation provides the answer. Intuition is nothing more and nothing lessthan recognition.”

This strong statement reduces the apparent magic of intuition to theeveryday experience of memory. We marvel at the story of the firefighter

(PDF) Thinking, fast and slow by daniel kahneman - PDFSLIDE.US (230)

who has a sudden urge to escape a burning house just before it collapses,because the firefighter knows the danger intuitively, “without knowing howhe knows.” However, we also do not know how we immediately know that aperson we see as we enter a room is our friend Peter. The moral ofSimon’s remark is that the mystery of knowing without knowing is not adistinctive feature of intuition; it is the norm of mental life.

Acquiring Skill

How does the information that supports intuition get “stored in memory”?Certain types of intuitions are acquired very quickly. We have inheritedfrom our ancestors a great facility to learn when to be afraid. Indeed, oneexperience is often sufficient to establish a long-term aversion and fear.Many of us have the visceral memory of a single dubious dish tto hat stillleaves us vaguely reluctant to return to a restaurant. All of us tense up whenwe approach a spot in which an unpleasant event occurred, even whenthere is no reason to expect it to happen again. For me, one such place isthe ramp leading to the San Francisco airport, where years ago a driver inthe throes of road rage followed me from the freeway, rolled down hiswindow, and hurled obscenities at me. I never knew what caused hishatred, but I remember his voice whenever I reach that point on my way tothe airport.

My memory of the airport incident is conscious and it fully explains theemotion that comes with it. On many occasions, however, you may feeluneasy in a particular place or when someone uses a particular turn ofphrase without having a conscious memory of the triggering event. Inhindsight, you will label that unease an intuition if it is followed by a badexperience. This mode of emotional learning is closely related to whathappened in Pavlov’s famous conditioning experiments, in which the dogslearned to recognize the sound of the bell as a signal that food wascoming. What Pavlov’s dogs learned can be described as a learned hope.Learned fears are even more easily acquired.

Fear can also be learned—quite easily, in fact—by words rather than byexperience. The fireman who had the “sixth sense” of danger had certainlyhad many occasions to discuss and think about types of fires he was notinvolved in, and to rehearse in his mind what the cues might be and how heshould react. As I remember from experience, a young platooncommander with no experience of combat will tense up while leadingtroops through a narrowing ravine, because he was taught to identify theterrain as favoring an ambush. Little repetition is needed for learning.

Emotional learning may be quick, but what we consider as “expertise”

(PDF) Thinking, fast and slow by daniel kahneman - PDFSLIDE.US (231)

usually takes a long time to develop. The acquisition of expertise incomplex tasks such as high-level chess, professional basketball, orfirefighting is intricate and slow because expertise in a domain is not asingle skill but rather a large collection of miniskills. Chess is a goodexample. An expert player can understand a complex position at a glance,but it takes years to develop that level of ability. Studies of chess mastershave shown that at least 10,000 hours of dedicated practice (about 6 yearsof playing chess 5 hours a day) are required to attain the highest levels ofperformance. During those hours of intense concentration, a serious chessplayer becomes familiar with thousands of configurations, each consistingof an arrangement of related pieces that can threaten or defend eachother.

Learning high-level chess can be compared to learning to read. A firstgrader works hard at recognizing individual letters and assembling theminto syllables and words, but a good adult reader perceives entire clauses.An expert reader has also acquired the ability to assemble familiarelements in a new pattern and can quickly “recognize” and correctlypronounce a word that she has never seen before. In chess, recurrentpatterns of interacting pieces play the role of letters, and a chess positionis a long word or a sentence.

A skilled reader who sees it for the first time will be able to read theopening stanza of Lewis Carroll’s “Jabberwocky” with perfect rhythm andintonation, as well as pleasure:

’Twas brillig, and the slithy tovesDid gyre and gimble in the wabe:All mimsy were the borogoves,And the mome raths outgrabe.

Acquiring expertise in chess is harder and slower than learning to readbecause there are many more letters in the “alphabet” of chess andbecause the “words” consist of many letters. After thousands of hours ofpractice, however, chess masters are able to read a chess situation at aglance. The few moves that come to their mind are almost always strongand sometimes creative. They can deal with a “word” they have neverencountered, and they can find a new way to interpret a familiar one.

The Environment of Skill

(PDF) Thinking, fast and slow by daniel kahneman - PDFSLIDE.US (232)

Klein and I quickly found that we agreed both on the nature of intuitive skilland on how it is acquired. We still needed to agree on our key question:When can you trust a self-confident professional who claims to have anintuition?

We eventually concluded that our disagreement was due in part to thefact that we had different experts in mind. Klein had spent much time withfireground commanders, clinical nurses, and other professionals who havereal expertise. I had spent more time thinking about clinicians, stockpickers, and political scientists trying to make unsupportable long-termforecasts. Not surprisingly, his default attitude was trust and respect; minewas skepticism. He was more willing to trust experts who claim an intuitionbecause, as he told me, true experts know the limits of their knowledge. Iargued that there are many pseudo-experts who have no idea that they donot know what they are doing (the illusion of validity), and that as a generalproposition subjective confidence is commonly too high and oftenuninformative.

Earlier I traced people’s confidence in a belief to two relatedimpressions: cognitive ease and coherence. We are confident when thestory we tell ourselves comes easily to mind, with no contradiction and nocompeting scenario. But ease and coherence do not guarantee that abelief held with confidence is true. The associative machine is set tosuppress doubt and to evoke ideas and information that are compatiblewith the currently dominant story. A mind that follows WY SIATI will achievehigh confidence much too easily by ignoring what it does not know. It istherefore not surprising that many of us are prone to have high confidencein unfounded intuitions. Klein and I eventually agreed on an importantprinciple: the confidence that people have in their intuitions is not a reliableguide to their validity. In other words, do not trust anyone—includingyourself—to tell you how much you should trust their judgment.

If subjective confidence is not to be trusted, how can we evaluate theprobable validity of an intuitive judgment? When do judgments reflect trueexpertise? When do they display an illusion of validity? The answer comesfrom the two basic conditions for acquiring a skill:

an environment that is sufficiently regular to be predictablean opportunity to learn these regularities through prolonged practice

When both these conditions are satisfied, intuitions are likely to be skilled.Chess is an extreme example of a regular environment, but bridge and

(PDF) Thinking, fast and slow by daniel kahneman - PDFSLIDE.US (233)

poker also provide robust statistical regularities that can support skill.Physicians, nurses, athletes, and firefighters also face complex butfundamentally orderly situations. The accurate intuitions that Gary Klein hasdescribed are due to highly valid cues that es the expert’s System 1 haslearned to use, even if System 2 has not learned to name them. In contrast,stock pickers and political scientists who make long-term forecastsoperate in a zero-validity environment. Their failures reflect the basicunpredictability of the events that they try to forecast.

Some environments are worse than irregular. Robin Hogarth described“wicked” environments, in which professionals are likely to learn the wronglessons from experience. He borrows from Lewis Thomas the example ofa physician in the early twentieth century who often had intuitions aboutpatients who were about to develop typhoid. Unfortunately, he tested hishunch by palpating the patient’s tongue, without washing his handsbetween patients. When patient after patient became ill, the physiciandeveloped a sense of clinical infallibility. His predictions were accurate—but not because he was exercising professional intuition!

Meehl’s clinicians were not inept and their failure was not due to lack oftalent. They performed poorly because they were assigned tasks that didnot have a simple solution. The clinicians’ predicament was less extremethan the zero-validity environment of long-term political forecasting, but theyoperated in low-validity situations that did not allow high accuracy. Weknow this to be the case because the best statistical algorithms, althoughmore accurate than human judges, were never very accurate. Indeed, thestudies by Meehl and his followers never produced a “smoking gun”demonstration, a case in which clinicians completely missed a highly validcue that the algorithm detected. An extreme failure of this kind is unlikelybecause human learning is normally efficient. If a strong predictive cueexists, human observers will find it, given a decent opportunity to do so.Statistical algorithms greatly outdo humans in noisy environments for tworeasons: they are more likely than human judges to detect weakly validcues and much more likely to maintain a modest level of accuracy by usingsuch cues consistently.

It is wrong to blame anyone for failing to forecast accurately in anunpredictable world. However, it seems fair to blame professionals forbelieving they can succeed in an impossible task. Claims for correctintuitions in an unpredictable situation are self-delusional at best,sometimes worse. In the absence of valid cues, intuitive “hits” are dueeither to luck or to lies. If you find this conclusion surprising, you still have alingering belief that intuition is magic. Remember this rule: intuition cannot

(PDF) Thinking, fast and slow by daniel kahneman - PDFSLIDE.US (234)

be trusted in the absence of stable regularities in the environment.

Feedback and Practice

Some regularities in the environment are easier to discover and apply thanothers. Think of how you developed your style of using the brakes on yourcar. As you were mastering the skill of taking curves, you gradually learnedwhen to let go of the accelerator and when and how hard to use the brakes.Curves differ, and the variability you experienced while learning ensuresthat you are now ready to brake at the right time and strength for any curveyou encounter. The conditions for learning this skill are ideal, because youreceive immediate and unambiguous feedback every time you go arounda bend: the mild reward of a comfortable turn or the mild punishment ofsome difficulty in handling the car if you brake either too hard or not quitehard enough. The situations that face a harbor pilot maneuvering largeships are no less regular, but skill is much more difficult to acquire by sheerexperience because of the long delay between actions and theirmanoticeable outcomes. Whether professionals have a chance to developintuitive expertise depends essentially on the quality and speed offeedback, as well as on sufficient opportunity to practice.

Expertise is not a single skill; it is a collection of skills, and the sameprofessional may be highly expert in some of the tasks in her domain whileremaining a novice in others. By the time chess players become experts,they have “seen everything” (or almost everything), but chess is anexception in this regard. Surgeons can be much more proficient in someoperations than in others. Furthermore, some aspects of anyprofessional’s tasks are much easier to learn than others.Psychotherapists have many opportunities to observe the immediatereactions of patients to what they say. The feedback enables them todevelop the intuitive skill to find the words and the tone that will calm anger,forge confidence, or focus the patient’s attention. On the other hand,therapists do not have a chance to identify which general treatmentapproach is most suitable for different patients. The feedback they receivefrom their patients’ long-term outcomes is sparse, delayed, or (usually)nonexistent, and in any case too ambiguous to support learning fromexperience.

Among medical specialties, anesthesiologists benefit from goodfeedback, because the effects of their actions are likely to be quicklyevident. In contrast, radiologists obtain little information about the accuracyof the diagnoses they make and about the pathologies they fail to detect.Anesthesiologists are therefore in a better position to develop useful

(PDF) Thinking, fast and slow by daniel kahneman - PDFSLIDE.US (235)

intuitive skills. If an anesthesiologist says, “I have a feeling something iswrong,” everyone in the operating room should be prepared for anemergency.

Here again, as in the case of subjective confidence, the experts may notknow the limits of their expertise. An experienced psychotherapist knowsthat she is skilled in working out what is going on in her patient’s mind andthat she has good intuitions about what the patient will say next. It istempting for her to conclude that she can also anticipate how well thepatient will do next year, but this conclusion is not equally justified. Short-term anticipation and long-term forecasting are different tasks, and thetherapist has had adequate opportunity to learn one but not the other.Similarly, a financial expert may have skills in many aspects of his tradebut not in picking stocks, and an expert in the Middle East knows manythings but not the future. The clinical psychologist, the stock picker, and thepundit do have intuitive skills in some of their tasks, but they have notlearned to identify the situations and the tasks in which intuition will betraythem. The unrecognized limits of professional skill help explain why expertsare often overconfident.

Evaluating Validity

At the end of our journey, Gary Klein and I agreed on a general answer toour initial question: When can you trust an experienced professional whoclaims to have an intuition? Our conclusion was that for the most part it ispossible to distinguish intuitions that are likely to be valid from those thatare likely to be bogus. As in the judgment of whether a work of art isgenuine or a fake, you will usually do better by focusing on its provenancethan by looking at the piece itself. If the environment is sufficiently regularand if the judge has had a chance to learn its regularities, the associativemachinery will recognize situations and generate quick and accuratepredictions and decisions. You can trust someone’s intuitions if theseconditions are met.

Unfortunately, associativentu memory also generates subjectivelycompelling intuitions that are false. Anyone who has watched the chessprogress of a talented youngster knows well that skill does not becomeperfect all at once, and that on the way to near perfection some mistakesare made with great confidence. When evaluating expert intuition youshould always consider whether there was an adequate opportunity tolearn the cues, even in a regular environment.

In a less regular, or low-validity, environment, the heuristics of judgmentare invoked. System 1 is often able to produce quick answers to difficult

(PDF) Thinking, fast and slow by daniel kahneman - PDFSLIDE.US (236)

questions by substitution, creating coherence where there is none. Thequestion that is answered is not the one that was intended, but the answeris produced quickly and may be sufficiently plausible to pass the lax andlenient review of System 2. You may want to forecast the commercial futureof a company, for example, and believe that this is what you are judging,while in fact your evaluation is dominated by your impressions of theenergy and competence of its current executives. Because substitutionoccurs automatically, you often do not know the origin of a judgment thatyou (your System 2) endorse and adopt. If it is the only one that comes tomind, it may be subjectively undistinguishable from valid judgments thatyou make with expert confidence. This is why subjective confidence is nota good diagnostic of accuracy: judgments that answer the wrong questioncan also be made with high confidence.

You may be asking, Why didn’t Gary Klein and I come up immediatelywith the idea of evaluating an expert’s intuition by assessing the regularityof the environment and the expert’s learning history—mostly setting asidethe expert’s confidence? And what did we think the answer could be?These are good questions because the contours of the solution wereapparent from the beginning. We knew at the outset that firegroundcommanders and pediatric nurses would end up on one side of theboundary of valid intuitions and that the specialties studied by Meehl wouldbe on the other, along with stock pickers and pundits.

It is difficult to reconstruct what it was that took us years, long hours ofdiscussion, endless exchanges of draft s and hundreds of e-mailsnegotiating over words, and more than once almost giving up. But this iswhat always happens when a project ends reasonably well: once youunderstand the main conclusion, it seems it was always obvious.

As the title of our article suggests, Klein and I disagreed less than wehad expected and accepted joint solutions of almost all the substantiveissues that were raised. However, we also found that our early differenceswere more than an intellectual disagreement. We had different attitudes,emotions, and tastes, and those changed remarkably little over the years.This is most obvious in the facts that we find amusing and interesting. Kleinstill winces when the word bias is mentioned, and he still enjoys stories inwhich algorithms or formal procedures lead to obviously absurd decisions.I tend to view the occasional failures of algorithms as opportunities toimprove them. On the other hand, I find more pleasure than Klein does inthe come-uppance of arrogant experts who claim intuitive powers in zero-validity situations. In the long run, however, finding as much intellectualagreement as we did is surely more important than the persistentemotional differences that remained.

(PDF) Thinking, fast and slow by daniel kahneman - PDFSLIDE.US (237)

Speaking of Expert Intuition

“How much expertise does she have in this particular task? Howmuch practice has she had?”

“Does he really believe that the environment of start-ups issufficiently regular to justify an intuition that goes against the baserates?”

“She is very confident in her decision, but subjective confidenceis a poor index of the accuracy of a judgment.”

“Did he really have an opportunity to learn? How quick and howclear was the feedback he received on his judgments?”

(PDF) Thinking, fast and slow by daniel kahneman - PDFSLIDE.US (238)

The Outside View

A few years after my collaboration with Amos began, I convinced someofficials in the Israeli Ministry of Education of the need for a curriculum toteach judgment and decision making in high schools. The team that Iassembled to design the curriculum and write a textbook for it includedseveral experienced teachers, some of my psychology students, andSeymour Fox, then dean of the Hebrew University’s School of Education,who was an expert in curriculum development.

After meeting every Friday afternoon for about a year, we hadconstructed a detailed outline of the syllabus, had written a couple ofchapters, and had run a few sample lessons in the classroom. We all feltthat we had made good progress. One day, as we were discussingprocedures for estimating uncertain quantities, the idea of conducting anexercise occurred to me. I asked everyone to write down an estimate ofhow long it would take us to submit a finished draft of the textbook to theMinistry of Education. I was following a procedure that we already plannedto incorporate into our curriculum: the proper way to elicit information froma group is not by starting with a public discussion but by confidentiallycollecting each person’s judgment. This procedure makes better use of theknowledge available to members of the group than the common practice ofopen discussion. I collected the estimates and jotted the results on theblackboard. They were narrowly centered around two years; the low endwas one and a half, the high end two and a half years.

Then I had another idea. I turned to Seymour, our curriculum expert, andasked whether he could think of other teams similar to ours that haddeveloped a curriculum from scratch. This was a time when severalpedagogical innovations like “new math” had been introduced, andSeymour said he could think of quite a few. I then asked whether he knewthe history of these teams in some detail, and it turned out that he wasfamiliar with several. I asked him to think of these teams when they hadmade as much progress as we had. How long, from that point, did it takethem to finish their textbook projects?

He fell silent. When he finally spoke, it seemed to me that he wasblushing, embarrassed by his own answer: “You know, I never realized thisbefore, but in fact not all the teams at a stage comparable to ours ever didcomplete their task. A substantial fraction of the teams ended up failing tofinish the job.”

This was worrisome; we had never considered the possibility that wemight fail. My anxiety rising, I asked how large he estimated that fractionwas. Rw l� sidering t20;About 40%,” he answered. By now, a pall of gloom

(PDF) Thinking, fast and slow by daniel kahneman - PDFSLIDE.US (239)

was falling over the room. The next question was obvious: “Those whofinished,” I asked. “How long did it take them?” “I cannot think of any groupthat finished in less than seven years,” he replied, “nor any that took morethan ten.”

I grasped at a straw: “When you compare our skills and resources tothose of the other groups, how good are we? How would you rank us incomparison with these teams?” Seymour did not hesitate long this time.“We’re below average,” he said, “but not by much.” This came as acomplete surprise to all of us—including Seymour, whose prior estimatehad been well within the optimistic consensus of the group. Until Iprompted him, there was no connection in his mind between hisknowledge of the history of other teams and his forecast of our future.

Our state of mind when we heard Seymour is not well described bystating what we “knew.” Surely all of us “knew” that a minimum of sevenyears and a 40% chance of failure was a more plausible forecast of thefate of our project than the numbers we had written on our slips of paper afew minutes earlier. But we did not acknowledge what we knew. The newforecast still seemed unreal, because we could not imagine how it couldtake so long to finish a project that looked so manageable. No crystal ballwas available to tell us the strange sequence of unlikely events that were inour future. All we could see was a reasonable plan that should produce abook in about two years, conflicting with statistics indicating that otherteams had failed or had taken an absurdly long time to complete theirmission. What we had heard was base-rate information, from which weshould have inferred a causal story: if so many teams failed, and if thosethat succeeded took so long, writing a curriculum was surely much harderthan we had thought. But such an inference would have conflicted with ourdirect experience of the good progress we had been making. Thestatistics that Seymour provided were treated as base rates normally are—noted and promptly set aside.

We should have quit that day. None of us was willing to invest six moreyears of work in a project with a 40% chance of failure. Although we musthave sensed that persevering was not reasonable, the warning did notprovide an immediately compelling reason to quit. After a few minutes ofdesultory debate, we gathered ourselves together and carried on as ifnothing had happened. The book was eventually completed eight(!) yearslater. By that time I was no longer living in Israel and had long since ceasedto be part of the team, which completed the task after many unpredictablevicissitudes. The initial enthusiasm for the idea in the Ministry of Educationhad waned by the time the text was delivered and it was never used.

This embarrassing episode remains one of the most instructiveexperiences of my professional life. I eventually learned three lessons from

(PDF) Thinking, fast and slow by daniel kahneman - PDFSLIDE.US (240)

it. The first was immediately apparent: I had stumbled onto a distinctionbetween two profoundly different approaches to forecasting, which Amosand I later labeled the inside view and the outside view. The second lessonwas that our initial forecasts of about two years for the completion of theproject exhibited a planning fallacy. Our estimates were closer to a best-case scenario than to a realistic assessment. I was slower to accept thethird lesson, which I call irrational perseverance: the folly we displayed thatday in failing to abandon the project. Facing a choice, we gave uprationality rather than give up the enterprise.

Drawn to the Inside ViewOn that long-ago Friday, our curriculum expert made two judgments aboutthe same problem and arrived at very different answers. The inside view isthe one that all of us, including Seymour, spontaneously adopted to assessthe future of our project. We focused on our specific circumstances andsearched for evidence in our own experiences. We had a sketchy plan: weknew how many chapters we were going to write, and we had an idea ofhow long it had taken us to write the two that we had already done. Themore cautious among us probably added a few months to their estimateas a margin of error.

Extrapolating was a mistake. We were forecasting based on theinformation in front of us—WYSIATI—but the chapters we wrote first wereprobably easier than others, and our commitment to the project wasprobably then at its peak. But the main problem was that we failed to allowfor what Donald Rumsfeld famously called the “unknown unknowns.” Therewas no way for us to foresee, that day, the succession of events that wouldcause the project to drag out for so long. The divorces, the illnesses, thecrises of coordination with bureaucracies that delayed the work could notbe anticipated. Such events not only cause the writing of chapters to slowdown, they also produce long periods during which little or no progress ismade at all. The same must have been true, of course, for the other teamsthat Seymour knew about. The members of those teams were also unableto imagine the events that would cause them to spend seven years tofinish, or ultimately fail to finish, a project that they evidently had thoughtwas very feasible. Like us, they did not know the odds they were facing.There are many ways for any plan to fail, and although most of them are tooimprobable to be anticipated, the likelihood that something will go wrongin a big project is high.

The second question I asked Seymour directed his attention away fromus and toward a class of similar cases. Seymour estimated the base rateof success in that reference class: 40% failure and seven to ten years for

(PDF) Thinking, fast and slow by daniel kahneman - PDFSLIDE.US (241)

completion. His informal survey was surely not up to scientific standards ofevidence, but it provided a reasonable basis for a baseline prediction: theprediction you make about a case if you know nothing except the categoryto which it belongs. As we saw earlier, the baseline prediction should bethe anchor for further adjustments. If you are asked to guess the height of awoman about whom you know only that she lives in New York City, yourbaseline prediction is your best guess of the average height of women inthe city. If you are now given case-specific information, for example that thewoman’s son is the starting center of his high school basketball team, youwill adjust your estimate away from the mean in the appropriate direction.Seymour’s comparison of our team to others suggested that the forecastof our outcome was slightly worse than the baseline prediction, which wasalready grim.

The spectacular accuracy of the outside-view forecast in our problemwas surely a fluke and should not count as evidence for the validity of theoutside view. The argument for the outside view should be made ongeneral grounds: if the reference class is properly chosen, the outside viewwill give an indication of where the ballpark is, and it may suggest, as it didin our case, that the inside-view forecasts are not even close to it.

For a psychologist, the discrepancy between Seymour’s two judgmentsis striking. He had in his head all the knowledge required to estimate thestatistics of an appropriate reference class, but he reached his initialestimate without ever using that knowledge. Seymour’s forecast from hisinsidethaa view was not an adjustment from the baseline prediction, whichhad not come to his mind. It was based on the particular circumstances ofour efforts. Like the participants in the Tom W experiment, Seymour knewthe relevant base rate but did not think of applying it.

Unlike Seymour, the rest of us did not have access to the outside viewand could not have produced a reasonable baseline prediction. It isnoteworthy, however, that we did not feel we needed information aboutother teams to make our guesses. My request for the outside viewsurprised all of us, including me! This is a common pattern: people whohave information about an individual case rarely feel the need to know thestatistics of the class to which the case belongs.

When we were eventually exposed to the outside view, we collectivelyignored it. We can recognize what happened to us; it is similar to theexperiment that suggested the futility of teaching psychology. When theymade predictions about individual cases about which they had a littleinformation (a brief and bland interview), Nisbett and Borgida’s studentscompletely neglected the global results they had just learned. “Pallid”statistical information is routinely discarded when it is incompatible with

(PDF) Thinking, fast and slow by daniel kahneman - PDFSLIDE.US (242)

one’s personal impressions of a case. In the competition with the insideview, the outside view doesn’t stand a chance.

The preference for the inside view sometimes carries moral overtones. Ionce asked my cousin, a distinguished lawyer, a question about areference class: “What is the probability of the defendant winning in caseslike this one?” His sharp answer that “every case is unique” wasaccompanied by a look that made it clear he found my questioninappropriate and superficial. A proud emphasis on the uniqueness ofcases is also common in medicine, in spite of recent advances inevidence-based medicine that point the other way. Medical statistics andbaseline predictions come up with increasing frequency in conversationsbetween patients and physicians. However, the remaining ambivalenceabout the outside view in the medical profession is expressed in concernsabout the impersonality of procedures that are guided by statistics andchecklists.

The Planning Fallacy

In light of both the outside-view forecast and the eventual outcome, theoriginal estimates we made that Friday afternoon appear almostdelusional. This should not come as a surprise: overly optimistic forecastsof the outcome of projects are found everywhere. Amos and I coined theterm planning fallacy to describe plans and forecasts that

are unrealistically close to best-case scenarioscould be improved by consulting the statistics of similar cases

Examples of the planning fallacy abound in the experiences ofindividuals, governments, and businesses. The list of horror stories isendless.

In July 1997, the proposed new Scottish Parliament building inEdinburgh was estimated to cost up to £40 million. By June 1999,the budget for the building was £109 million. In April 2000, legislatorsimposed a £195 million “cap on costs.” By November 2001, theydemanded an estimate of “final cost,” which was set at £241 million.That estimated final cost rose twice in 2002, ending the year at

(PDF) Thinking, fast and slow by daniel kahneman - PDFSLIDE.US (243)

£294.6 million. It rose three times more in 2003, reaching £375.8million by June. The building was finally comanspleted in 2004 at anultimate cost of roughly £431 million.A 2005 study examined rail projects undertaken worldwide between1969 and 1998. In more than 90% of the cases, the number ofpassengers projected to use the system was overestimated. Eventhough these passenger shortfalls were widely publicized, forecastsdid not improve over those thirty years; on average, plannersoverestimated how many people would use the new rail projects by106%, and the average cost overrun was 45%. As more evidenceaccumulated, the experts did not become more reliant on it.In 2002, a survey of American homeowners who had remodeled theirkitchens found that, on average, they had expected the job to cost$18,658; in fact, they ended up paying an average of $38,769.

The optimism of planners and decision makers is not the only cause ofoverruns. Contractors of kitchen renovations and of weapon systemsreadily admit (though not to their clients) that they routinely make most oftheir profit on additions to the original plan. The failures of forecasting inthese cases reflect the customers’ inability to imagine how much theirwishes will escalate over time. They end up paying much more than theywould if they had made a realistic plan and stuck to it.

Errors in the initial budget are not always innocent. The authors ofunrealistic plans are often driven by the desire to get the plan approved—whether by their superiors or by a client—supported by the knowledge thatprojects are rarely abandoned unfinished merely because of overruns incosts or completion times. In such cases, the greatest responsibility foravoiding the planning fallacy lies with the decision makers who approvethe plan. If they do not recognize the need for an outside view, they commita planning fallacy.

Mitigating the Planning Fallacy

The diagnosis of and the remedy for the planning fallacy have not changedsince that Friday afternoon, but the implementation of the idea has come along way. The renowned Danish planning expert Bent Flyvbjerg, now atOxford University, offered a forceful summary:

The prevalent tendency to underweight or ignore distributionalinformation is perhaps the major source of error in forecasting.Planners should therefore make every effort to frame the

(PDF) Thinking, fast and slow by daniel kahneman - PDFSLIDE.US (244)

forecasting problem so as to facilitate utilizing all thedistributional information that is available.

This may be considered the single most important piece of adviceregarding how to increase accuracy in forecasting through improvedmethods. Using such distributional information from other ventures similarto that being forecasted is called taking an “outside view” and is the cure tothe planning fallacy.

The treatment for the planning fallacy has now acquired a technicalname, reference class forecasting, and Flyvbjerg has applied it totransportation projects in several countries. The outside view isimplemented by using a large database, which provides information onboth plans and outcomes for hundreds of projects all over the world, andcan be used to provide statistical information about the likely overruns ofcost and time, and about the likely underperformance of projects ofdifferent types.

The forecasting method that Flyvbjerg applies is similar to the practicesrecommended for overcoming base-rate neglect:

1. Identify an appropriate reference class (kitchen renovations, largerailway projects, etc.).

2. Obtain the statistics of the reference class (in terms of cost per mileof railway, or of the percentage by which expenditures exceededbudget). Use the statistics to generate a baseline prediction.

3. Use specific information about the case to adjust the baselineprediction, if there are particular reasons to expect the optimisticbias to be more or less pronounced in this project than in others ofthe same type.

Flyvbjerg’s analyses are intended to guide the authorities that commissionpublic projects, by providing the statistics of overruns in similar projects.Decision makers need a realistic assessment of the costs and benefits ofa proposal before making the final decision to approve it. They may alsowish to estimate the budget reserve that they need in anticipation ofoverruns, although such precautions often become self-fulfillingprophecies. As one official told Flyvbjerg, “A budget reserve is tocontractors as red meat is to lions, and they will devour it.”

Organizations face the challenge of controlling the tendency ofexecutives competing for resources to present overly optimistic plans. Awell-run organization will reward planners for precise execution and

(PDF) Thinking, fast and slow by daniel kahneman - PDFSLIDE.US (245)

penalize them for failing to anticipate difficulties, and for failing to allow fordifficulties that they could not have anticipated—the unknown unknowns.

Decisions and Errors

That Friday afternoon occurred more than thirty years ago. I often thoughtabout it and mentioned it in lectures several times each year. Some of myfriends got bored with the story, but I kept drawing new lessons from it.Almost fifteen years after I first reported on the planning fallacy with Amos, Ireturned to the topic with Dan Lovallo. Together we sketched a theory ofdecision making in which the optimistic bias is a significant source of risktaking. In the standard rational model of economics, people take risksbecause the odds are favorable—they accept some probability of a costlyfailure because the probability of success is sufficient. We proposed analternative idea.

When forecasting the outcomes of risky projects, executives too easilyfall victim to the planning fallacy. In its grip, they make decisions based ondelusional optimism rather than on a rational weighting of gains, losses,and probabilities. They overestimate benefits and underestimate costs.They spin scenarios of success while overlooking the potential formistakes and miscalculations. As a result, they pursue initiatives that areunlikely to come in on budget or on time or to deliver the expected returns—or even to be completed.

In this view, people often (but not always) take on risky projects becausethey are overly optimistic about the odds they face. I will return to this ideaseveral times in this book—it probably contributes to an explanation of whypeople litigate, why they start wars, and why they open small businesses.

Failing a Test

For many years, I thought that the main point of the curriculum story waswhat I had learned about my friend Seymour: that his best guess about thefuture of our project was not informed by what he knew about similarprojects. I came off quite well in my telling of the story, ir In which I had therole of clever questioner and astute psychologist. I only recently realizedthat I had actually played the roles of chief dunce and inept leader.

The project was my initiative, and it was therefore my responsibility toensure that it made sense and that major problems were properlydiscussed by the team, but I failed that test. My problem was no longer theplanning fallacy. I was cured of that fallacy as soon as I heard Seymour’sstatistical summary. If pressed, I would have said that our earlier estimates

(PDF) Thinking, fast and slow by daniel kahneman - PDFSLIDE.US (246)

had been absurdly optimistic. If pressed further, I would have admitted thatwe had started the project on faulty premises and that we should at leastconsider seriously the option of declaring defeat and going home. Butnobody pressed me and there was no discussion; we tacitly agreed to goon without an explicit forecast of how long the effort would last. This waseasy to do because we had not made such a forecast to begin with. If wehad had a reasonable baseline prediction when we started, we would nothave gone into it, but we had already invested a great deal of effort—aninstance of the sunk-cost fallacy, which we will look at more closely in thenext part of the book. It would have been embarrassing for us—especiallyfor me—to give up at that point, and there seemed to be no immediatereason to do so. It is easier to change directions in a crisis, but this wasnot a crisis, only some new facts about people we did not know. Theoutside view was much easier to ignore than bad news in our own effort. Ican best describe our state as a form of lethargy—an unwillingness to thinkabout what had happened. So we carried on. There was no further attemptat rational planning for the rest of the time I spent as a member of the team—a particularly troubling omission for a team dedicated to teachingrationality. I hope I am wiser today, and I have acquired a habit of lookingfor the outside view. But it will never be the natural thing to do.

Speaking of the Outside View

“He’s taking an inside view. He should forget about his own caseand look for what happened in other cases.”

“She is the victim of a planning fallacy. She’s assuming a best-case scenario, but there are too many different ways for the planto fail, and she cannot foresee them all.”

“Suppose you did not know a thing about this particular legalcase, only that it involves a malpractice claim by an individualagainst a surgeon. What would be your baseline prediction? Howmany of these cases succeed in court? How many settle? Whatare the amounts? Is the case we are discussing stronger orweaker than similar claims?”

“We are making an additional investment because we do not

(PDF) Thinking, fast and slow by daniel kahneman - PDFSLIDE.US (247)

want to admit failure. This is an instance of the sunk-cost fallacy.”

(PDF) Thinking, fast and slow by daniel kahneman - PDFSLIDE.US (248)

The Engine of Capitalism

The planning fallacy is only one of the manifestations of a pervasiveoptimistic bias. sid to adtions of aMost of us view the world as morebenign than it really is, our own attributes as more favorable than they trulyare, and the goals we adopt as more achievable than they are likely to be.We also tend to exaggerate our ability to forecast the future, which fostersoptimistic overconfidence. In terms of its consequences for decisions, theoptimistic bias may well be the most significant of the cognitive biases.Because optimistic bias can be both a blessing and a risk, you should beboth happy and wary if you are temperamentally optimistic.

Optimists

Optimism is normal, but some fortunate people are more optimistic thanthe rest of us. If you are genetically endowed with an optimistic bias, youhardly need to be told that you are a lucky person—you already feelfortunate. An optimistic attitude is largely inherited, and it is part of ageneral disposition for well-being, which may also include a preference forseeing the bright side of everything. If you were allowed one wish for yourchild, seriously consider wishing him or her optimism. Optimists arenormally cheerful and happy, and therefore popular; they are resilient inadapting to failures and hardships, their chances of clinical depression arereduced, their immune system is stronger, they take better care of theirhealth, they feel healthier than others and are in fact likely to live longer. Astudy of people who exaggerate their expected life span beyond actuarialpredictions showed that they work longer hours, are more optimistic abouttheir future income, are more likely to remarry after divorce (the classic“triumph of hope over experience”), and are more prone to bet onindividual stocks. Of course, the blessings of optimism are offered only toindividuals who are only mildly biased and who are able to “accentuate thepositive” without losing track of reality.

Optimistic individuals play a disproportionate role in shaping our lives.Their decisions make a difference; they are the inventors, theentrepreneurs, the political and military leaders—not average people. Theygot to where they are by seeking challenges and taking risks. They aretalented and they have been lucky, almost certainly luckier than theyacknowledge. They are probably optimistic by temperament; a survey offounders of small businesses concluded that entrepreneurs are moresanguine than midlevel managers about life in general. Their experiencesof success have confirmed their faith in their judgment and in their ability to

(PDF) Thinking, fast and slow by daniel kahneman - PDFSLIDE.US (249)

control events. Their self-confidence is reinforced by the admiration ofothers. This reasoning leads to a hypothesis: the people who have thegreatest influence on the lives of others are likely to be optimistic andoverconfident, and to take more risks than they realize.

The evidence suggests that an optimistic bias plays a role—sometimesthe dominant role—whenever individuals or institutions voluntarily take onsignificant risks. More often than not, risk takers underestimate the oddsthey face, and do invest sufficient effort to find out what the odds are.Because they misread the risks, optimistic entrepreneurs often believethey are prudent, even when they are not. Their confidence in their futuresuccess sustains a positive mood that helps them obtain resources fromothers, raise the morale of their employees, and enhance their prospectsof prevailing. When action is needed, optimism, even of the mildlydelusional variety, may be a good thing.

Entrepreneurial Delusions

The chances that a small business will thesurvive for five years in theUnited States are about 35%. But the individuals who open suchbusinesses do not believe that the statistics apply to them. A survey foundthat American entrepreneurs tend to believe they are in a promising line ofbusiness: their average estimate of the chances of success for “anybusiness like yours” was 60%—almost double the true value. The bias wasmore glaring when people assessed the odds of their own venture. Fully81% of the entrepreneurs put their personal odds of success at 7 out of 10or higher, and 33% said their chance of failing was zero.

The direction of the bias is not surprising. If you interviewed someonewho recently opened an Italian restaurant, you would not expect her to haveunderestimated her prospects for success or to have a poor view of herability as a restaurateur. But you must wonder: Would she still haveinvested money and time if she had made a reasonable effort to learn theodds—or, if she did learn the odds (60% of new restaurants are out ofbusiness after three years), paid attention to them? The idea of adoptingthe outside view probably didn’t occur to her.

One of the benefits of an optimistic temperament is that it encouragespersistence in the face of obstacles. But persistence can be costly. Animpressive series of studies by Thomas Åstebro sheds light on whathappens when optimists receive bad news. He drew his data from aCanadian organization—the Inventor’s Assistance Program—which

(PDF) Thinking, fast and slow by daniel kahneman - PDFSLIDE.US (250)

collects a small fee to provide inventors with an objective assessment ofthe commercial prospects of their idea. The evaluations rely on carefulratings of each invention on 37 criteria, including need for the product, costof production, and estimated trend of demand. The analysts summarizetheir ratings by a letter grade, where D and E predict failure—a predictionmade for over 70% of the inventions they review. The forecasts of failureare remarkably accurate: only 5 of 411 projects that were given the lowestgrade reached commercialization, and none was successful.

Discouraging news led about half of the inventors to quit after receivinga grade that unequivocally predicted failure. However, 47% of themcontinued development efforts even after being told that their project washopeless, and on average these persistent (or obstinate) individualsdoubled their initial losses before giving up. Significantly, persistence afterdiscouraging advice was relatively common among inventors who had ahigh score on a personality measure of optimism—on which inventorsgenerally scored higher than the general population. Overall, the return onprivate invention was small, “lower than the return on private equity and onhigh-risk securities.” More generally, the financial benefits of self-employment are mediocre: given the same qualifications, people achievehigher average returns by selling their skills to employers than by settingout on their own. The evidence suggests that optimism is widespread,stubborn, and costly.

Psychologists have confirmed that most people genuinely believe thatthey are superior to most others on most desirable traits—they are willingto bet small amounts of money on these beliefs in the laboratory. In themarket, of course, beliefs in one’s superiority have significantconsequences. Leaders of large businesses sometimes make huge betsin expensive mergers and acquisitions, acting on the mistaken belief thatthey can manage the assets of another company better than its currentowners do. The stock market commonly responds by downgrading thevalue of the acquiring firm, because experience has shown that efforts tointegrate large firms fail more often than they succeed. The misguidedacquisitions have been explained by a “hubris hypothesis”: the eivxecutives of the acquiring firm are simply less competent than they thinkthey are.

The economists Ulrike Malmendier and Geoffrey Tate identifiedoptimistic CEOs by the amount of company stock that they ownedpersonally and observed that highly optimistic leaders took excessiverisks. They assumed debt rather than issue equity and were more likelythan others to “overpay for target companies and undertake value-destroying mergers.” Remarkably, the stock of the acquiring companysuffered substantially more in mergers if the CEO was overly optimistic by

(PDF) Thinking, fast and slow by daniel kahneman - PDFSLIDE.US (251)

the authors’ measure. The stock market is apparently able to identifyoverconfident CEOs. This observation exonerates the CEOs from oneaccusation even as it convicts them of another: the leaders of enterpriseswho make unsound bets do not do so because they are betting with otherpeople’s money. On the contrary, they take greater risks when theypersonally have more at stake. The damage caused by overconfidentCEOs is compounded when the business press anoints them ascelebrities; the evidence indicates that prestigious press awards to theCEO are costly to stockholders. The authors write, “We find that firms withaward-winning CEOs subsequently underperform, in terms both of stockand of operating performance. At the same time, CEO compensationincreases, CEOs spend more time on activities outside the company suchas writing books and sitting on outside boards, and they are more likely toengage in earnings management.”

Many years ago, my wife and I were on vacation on Vancouver Island,looking for a place to stay. We found an attractive but deserted motel on alittle-traveled road in the middle of a forest. The owners were a charmingyoung couple who needed little prompting to tell us their story. They hadbeen schoolteachers in the province of Alberta; they had decided tochange their life and used their life savings to buy this motel, which hadbeen built a dozen years earlier. They told us without irony or self-consciousness that they had been able to buy it cheap, “because six orseven previous owners had failed to make a go of it.” They also told usabout plans to seek a loan to make the establishment more attractive bybuilding a restaurant next to it. They felt no need to explain why theyexpected to succeed where six or seven others had failed. A commonthread of boldness and optimism links businesspeople, from motel ownersto superstar CEOs.

The optimistic risk taking of entrepreneurs surely contributes to theeconomic dynamism of a capitalistic society, even if most risk takers endup disappointed. However, Marta Coelho of the London School ofEconomics has pointed out the difficult policy issues that arise whenfounders of small businesses ask the government to support them indecisions that are most likely to end badly. Should the government provideloans to would-be entrepreneurs who probably will bankrupt themselves ina few years? Many behavioral economists are comfortable with the“libertarian paternalistic” procedures that help people increase theirsavings rate beyond what they would do on their own. The question ofwhether and how government should support small business does not have

(PDF) Thinking, fast and slow by daniel kahneman - PDFSLIDE.US (252)

an equally satisfying answer.

Competition Neglect

It is tempting to explain entrepreneurial optimism by wishful thinking, butemotion is only part of the story. Cognitive biases play an important role,notably the System 1 feature WYSIATI.

We focus on our goal, anchor on our plan, and neglect relevant baserates, exposing ourselves to tnesehe planning fallacy.We focus on what we want to do and can do, neglecting the plansand skills of others.Both in explaining the past and in predicting the future, we focus onthe causal role of skill and neglect the role of luck. We are thereforeprone to an illusion of control.We focus on what we know and neglect what we do not know, whichmakes us overly confident in our beliefs.

The observation that “90% of drivers believe they are better thanaverage” is a well-established psychological finding that has become partof the culture, and it often comes up as a prime example of a more generalabove-average effect. However, the interpretation of the finding haschanged in recent years, from self-aggrandizement to a cognitive bias.Consider these two questions:

Are you a good driver?Are you better than average as a driver?

The first question is easy and the answer comes quickly: most drivers sayyes. The second question is much harder and for most respondents almostimpossible to answer seriously and correctly, because it requires anassessment of the average quality of drivers. At this point in the book itcomes as no surprise that people respond to a difficult question byanswering an easier one. They compare themselves to the averagewithout ever thinking about the average. The evidence for the cognitiveinterpretation of the above-average effect is that when people are askedabout a task they find difficult (for many of us this could be “Are you betterthan average in starting conversations with strangers?”), they readily ratethemselves as below average. The upshot is that people tend to be overly

(PDF) Thinking, fast and slow by daniel kahneman - PDFSLIDE.US (253)

optimistic about their relative standing on any activity in which they domoderately well.

I have had several occasions to ask founders and participants ininnovative start-ups a question: To what extent will the outcome of youreffort depend on what you do in your firm? This is evidently an easyquestion; the answer comes quickly and in my small sample it has neverbeen less than 80%. Even when they are not sure they will succeed, thesebold people think their fate is almost entirely in their own hands. They aresurely wrong: the outcome of a start-up depends as much on theachievements of its competitors and on changes in the market as on itsown efforts. However, WY SIATI plays its part, and entrepreneurs naturallyfocus on what they know best—their plans and actions and the mostimmediate threats and opportunities, such as the availability of funding.They know less about their competitors and therefore find it natural toimagine a future in which the competition plays little part.

Colin Camerer and Dan Lovallo, who coined the concept of competitionneglect, illustrated it with a quote from the then chairman of DisneyStudios. Asked why so many expensive big-budget movies are releasedon the same days (such as Memorial Day and Independence Day), hereplied:

Hubris. Hubris. If you only think about your own business, youthink, “I’ve got a good story department, I’ve got a goodmarketing department, we’re going to go out and do this.” Andyou don’t think that everybody else is thinking the same way. In agiven weekend in a year you’ll have five movies open, and there’scertainly not enough people to go around. re

The candid answer refers to hubris, but it displays no arrogance, noconceit of superiority to competing studios. The competition is simply notpart of the decision, in which a difficult question has again been replacedby an easier one. The question that needs an answer is this: Consideringwhat others will do, how many people will see our film? The question thestudio executives considered is simpler and refers to knowledge that ismost easily available to them: Do we have a good film and a goodorganization to market it? The familiar System 1 processes of WY SIATIand substitution produce both competition neglect and the above-averageeffect. The consequence of competition neglect is excess entry: morecompetitors enter the market than the market can profitably sustain, sotheir average outcome is a loss. The outcome is disappointing for thetypical entrant in the market, but the effect on the economy as a wholecould well be positive. In fact, Giovanni Dosi and Dan Lovallo call

(PDF) Thinking, fast and slow by daniel kahneman - PDFSLIDE.US (254)

entrepreneurial firms that fail but signal new markets to more qualifiedcompetitors “optimistic martyrs”—good for the economy but bad for theirinvestors.

Overconfidence

For a number of years, professors at Duke University conducted a surveyin which the chief financial officers of large corporations estimated thereturns of the Standard & Poor’s index over the following year. The Dukescholars collected 11,600 such forecasts and examined their accuracy.The conclusion was straightforward: financial officers of large corporationshad no clue about the short-term future of the stock market; the correlationbetween their estimates and the true value was slightly less than zero!When they said the market would go down, it was slightly more likely thannot that it would go up. These findings are not surprising. The truly badnews is that the CFOs did not appear to know that their forecasts wereworthless.

In addition to their best guess about S&P returns, the participantsprovided two other estimates: a value that they were 90% sure would betoo high, and one that they were 90% sure would be too low. The rangebetween the two values is called an “80% confidence interval” andoutcomes that fall outside the interval are labeled “surprises.” An individualwho sets confidence intervals on multiple occasions expects about 20% ofthe outcomes to be surprises. As frequently happens in such exercises,there were far too many surprises; their incidence was 67%, more than 3times higher than expected. This shows that CFOs were grosslyoverconfident about their ability to forecast the market. Overconfidence isanother manifestation of WYSIATI: when we estimate a quantity, we rely oninformation that comes to mind and construct a coherent story in which theestimate makes sense. Allowing for the information that does not come tomind—perhaps because one never knew it—is impossible.

The authors calculated the confidence intervals that would have reducedthe incidence of surprises to 20%. The results were striking. To maintainthe rate of surprises at the desired level, the CFOs should have said, yearafter year, “There is an 80% chance that the S&P return next year will bebetween –10% and +30%.” The confidence interval that properly reflectsthe CFOs’ knowledge (more precisely, their ignorance) is more than 4times wider than the intervals they actually stated.

Social psychology comes into the picture here, because the answer thata truthful CFO would offer is plainly ridiculous. A CFO who informs hiscolleagues that “th%">iere is a good chance that the S&P returns will be

(PDF) Thinking, fast and slow by daniel kahneman - PDFSLIDE.US (255)

between –10% and +30%” can expect to be laughed out of the room. Thewide confidence interval is a confession of ignorance, which is not sociallyacceptable for someone who is paid to be knowledgeable in financialmatters. Even if they knew how little they know, the executives would bepenalized for admitting it. President Truman famously asked for a “one-armed economist” who would take a clear stand; he was sick and tired ofeconomists who kept saying, “On the other hand…”

Organizations that take the word of overconfident experts can expectcostly consequences. The study of CFOs showed that those who weremost confident and optimistic about the S&P index were alsooverconfident and optimistic about the prospects of their own firm, whichwent on to take more risk than others. As Nassim Taleb has argued,inadequate appreciation of the uncertainty of the environment inevitablyleads economic agents to take risks they should avoid. However, optimismis highly valued, socially and in the market; people and firms reward theproviders of dangerously misleading information more than they rewardtruth tellers. One of the lessons of the financial crisis that led to the GreatRecession is that there are periods in which competition, among expertsand among organizations, creates powerful forces that favor a collectiveblindness to risk and uncertainty.

The social and economic pressures that favor overconfidence are notrestricted to financial forecasting. Other professionals must deal with thefact that an expert worthy of the name is expected to display highconfidence. Philip Tetlock observed that the most overconfident expertswere the most likely to be invited to strut their stuff in news shows.Overconfidence also appears to be endemic in medicine. A study ofpatients who died in the ICU compared autopsy results with the diagnosisthat physicians had provided while the patients were still alive. Physiciansalso reported their confidence. The result: “clinicians who were ‘completelycertain’ of the diagnosis antemortem were wrong 40% of the time.” Hereagain, expert overconfidence is encouraged by their clients: “Generally, itis considered a weakness and a sign of vulnerability for clinicians toappear unsure. Confidence is valued over uncertainty and there is aprevailing censure against disclosing uncertainty to patients.” Experts whoacknowledge the full extent of their ignorance may expect to be replacedby more confident competitors, who are better able to gain the trust ofclients. An unbiased appreciation of uncertainty is a cornerstone ofrationality—but it is not what people and organizations want. Extremeuncertainty is paralyzing under dangerous circumstances, and theadmission that one is merely guessing is especially unacceptable whenthe stakes are high. Acting on pretended knowledge is often the preferredsolution.

(PDF) Thinking, fast and slow by daniel kahneman - PDFSLIDE.US (256)

When they come together, the emotional, cognitive, and social factorsthat support exaggerated optimism are a heady brew, which sometimesleads people to take risks that they would avoid if they knew the odds.There is no evidence that risk takers in the economic domain have anunusual appetite for gambles on high stakes; they are merely less aware ofrisks than more timid people are. Dan Lovallo and I coined the phrase“bold forecasts and timid decisions” to describe the background of risktaking.

The effects of high optimism on decision making are, at best, a mixedblessing, but the contribution of optimism to good implementation iscertainly positive. The main benefit of optimism is resilience in the face ofsetbacks. According to Martin Seligman, the founder of potelsitivepsychology, an “optimistic explanation style” contributes to resilience bydefending one’s self-image. In essence, the optimistic style involves takingcredit for successes but little blame for failures. This style can be taught, atleast to some extent, and Seligman has documented the effects of trainingon various occupations that are characterized by a high rate of failures,such as cold-call sales of insurance (a common pursuit in pre-Internetdays). When one has just had a door slammed in one’s face by an angryhomemaker, the thought that “she was an awful woman” is clearly superiorto “I am an inept salesperson.” I have always believed that scientificresearch is another domain where a form of optimism is essential tosuccess: I have yet to meet a successful scientist who lacks the ability toexaggerate the importance of what he or she is doing, and I believe thatsomeone who lacks a delusional sense of significance will wilt in the faceof repeated experiences of multiple small failures and rare successes, thefate of most researchers.

The Premortem: A Partial Remedy

Can overconfident optimism be overcome by training? I am not optimistic.There have been numerous attempts to train people to state confidenceintervals that reflect the imprecision of their judgments, with only a fewreports of modest success. An often cited example is that geologists atRoyal Dutch Shell became less overconfident in their assessments ofpossible drilling sites after training with multiple past cases for which theoutcome was known. In other situations, overconfidence was mitigated (butnot eliminated) when judges were encouraged to consider competinghypotheses. However, overconfidence is a direct consequence of features

(PDF) Thinking, fast and slow by daniel kahneman - PDFSLIDE.US (257)

of System 1 that can be tamed—but not vanquished. The main obstacle isthat subjective confidence is determined by the coherence of the story onehas constructed, not by the quality and amount of the information thatsupports it.

Organizations may be better able to tame optimism and individuals thanindividuals are. The best idea for doing so was contributed by Gary Klein,my “adversarial collaborator” who generally defends intuitive decisionmaking against claims of bias and is typically hostile to algorithms. Helabels his proposal the premortem. The procedure is simple: when theorganization has almost come to an important decision but has not formallycommitted itself, Klein proposes gathering for a brief session a group ofindividuals who are knowledgeable about the decision. The premise of thesession is a short speech: “Imagine that we are a year into the future. Weimplemented the plan as it now exists. The outcome was a disaster.Please take 5 to 10 minutes to write a brief history of that disaster.”

Gary Klein’s idea of the premortem usually evokes immediateenthusiasm. After I described it casually at a session in Davos, someonebehind me muttered, “It was worth coming to Davos just for this!” (I laternoticed that the speaker was the CEO of a major internationalcorporation.) The premortem has two main advantages: it overcomes thegroupthink that affects many teams once a decision appears to have beenmade, and it unleashes the imagination of knowledgeable individuals in amuch-needed direction.

As a team converges on a decision—and especially when the leadertips her hand—public doubts about the wisdom of the planned move aregradually suppressed and eventually come to be treated as evidence offlawed loyalty to the team and its leaders. The suppression of doubtcontributes to overconfidence in a group where only supporters of thedecision have a v filepos-id="filepos726557"> nacea and does notprovide complete protection against nasty surprises, but it goes some waytoward reducing the damage of plans that are subject to the biases of WYSIATI and uncritical optimism.

Speaking of Optimism

“They have an illusion of control. They seriously underestimate theobstacles.”

“They seem to suffer from an acute case of competitor neglect.”

(PDF) Thinking, fast and slow by daniel kahneman - PDFSLIDE.US (258)

“This is a case of overconfidence. They seem to believe theyknow more than they actually do know.”

“We should conduct a premortem session. Someone may comeup with a threat we have neglected.”

(PDF) Thinking, fast and slow by daniel kahneman - PDFSLIDE.US (259)

Part 4

(PDF) Thinking, fast and slow by daniel kahneman - PDFSLIDE.US (260)

Choices

(PDF) Thinking, fast and slow by daniel kahneman - PDFSLIDE.US (261)

Bernoulli’s Errors

One day in the early 1970s, Amos handed me a mimeographed essay bya Swiss economist named Bruno Frey, which discussed the psychologicalassumptions of economic theory. I vividly remember the color of the cover:dark red. Bruno Frey barely recalls writing the piece, but I can still recite itsfirst sentence: “The agent of economic theory is rational, selfish, and histastes do not change.”

I was astonished. My economist colleagues worked in the building nextdoor, but I had not appreciated the profound difference between ourintellectual worlds. To a psychologist, it is self-evident that people areneither fully rational nor completely selfish, and that their tastes areanything but stable. Our two disciplines seemed to be studying differentspecies, which the behavioral economist Richard Thaler later dubbedEcons and Humans.

Unlike Econs, the Humans that psychologists know have a System 1.Their view of the world is limited by the information that is available at agiven moment (WYSIATI), and therefore they cannot be as consistent andlogical as Econs. They are sometimes generous and often willing tocontribute to the group to which they are attached. And they often have littleidea of what they will like next year or even tomorrow. Here was anopportunity for an interesting conversation across the boundaries of thedisciplines. I did not anticipate that my career would be defined by thatconversation.

Soon after he showed me Frey’s article, Amos suggested that we makethe study of decision making our next project. I knew next to nothing aboutthe topic, but Amos was an expert and a star of the field, and heMathematical Psychology, and he directed me to a few chapters that hethought would be a good introduction.

I soon learned that our subject matter would be people’s attitudes torisky options and that we would seek to answer a specific question: Whatrules govern people’s choices between different simple gambles andbetween gambles and sure things?

Simple gambles (such as “40% chance to win $300”) are to students ofdecision making what the fruit fly is to geneticists. Choices between suchgambles provide a simple model that shares important features with themore complex decisions that researchers actually aim to understand.Gambles represent the fact that the consequences of choices are nevercertain. Even ostensibly sure outcomes are uncertain: when you sign thecontract to buy an apartment, you do not know the price at which you latermay have to sell it, nor do you know that your neighbor’s son will soon take

(PDF) Thinking, fast and slow by daniel kahneman - PDFSLIDE.US (262)

up the tuba. Every significant choice we make in life comes with someuncertainty—which is why students of decision making hope that some ofthe lessons learned in the model situation will be applicable to moreinteresting everyday problems. But of course the main reason that decisiontheorists study simple gambles is that this is what other decision theoristsdo.

The field had a theory, expected utility theory, which was the foundationof the rational-agent model and is to this day the most important theory inthe social sciences. Expected utility theory was not intended as apsychological model; it was a logic of choice, based on elementary rules(axioms) of rationality. Consider this example:

If you prefer an apple to a banana,thenyou also prefer a 10% chance to win an apple to a 10% chanceto win a banana.

The apple and the banana stand for any objects of choice (includinggambles), and the 10% chance stands for any probability. Themathematician John von Neumann, one of the giant intellectual figures ofthe twentieth century, and the economist Oskar Morgenstern had derivedtheir theory of rational choice between gambles from a few axioms.Economists adopted expected utility theory in a dual role: as a logic thatprescribes how decisions should be made, and as a description of howEcons make choices. Amos and I were psychologists, however, and weset out to understand how Humans actually make risky choices, withoutassuming anything about their rationality.

We maintained our routine of spending many hours each day inconversation, sometimes in our offices, sometimes at restaurants, often onlong walks through the quiet streets of beautiful Jerusalem. As we haddone when we studied judgment, we engaged in a careful examination ofour own intuitive preferences. We spent our time inventing simple decisionproblems and asking ourselves how we would choose. For example:

Which do you prefer?A. Toss a coin. If it comes up heads you win $100, and if it comesup tails you win nothing.B. Get $46 for sure.

We were not trying to figure out the mos BineithWe t rational oradvantageous choice; we wanted to find the intuitive choice, the one thatappeared immediately tempting. We almost always selected the same

(PDF) Thinking, fast and slow by daniel kahneman - PDFSLIDE.US (263)

option. In this example, both of us would have picked the sure thing, andyou probably would do the same. When we confidently agreed on a choice,we believed—almost always correctly, as it turned out—that most peoplewould share our preference, and we moved on as if we had solid evidence.We knew, of course, that we would need to verify our hunches later, but byplaying the roles of both experimenters and subjects we were able to movequickly.

Five years after we began our study of gambles, we finally completed anessay that we titled “Prospect Theory: An Analysis of Decision under Risk.”Our theory was closely modeled on utility theory but departed from it infundamental ways. Most important, our model was purely descriptive, andits goal was to document and explain systematic violations of the axiomsof rationality in choices between gambles. We submitted our essay toEconometrica, a journal that publishes significant theoretical articles ineconomics and in decision theory. The choice of venue turned out to beimportant; if we had published the identical paper in a psychologicaljournal, it would likely have had little impact on economics. However, ourdecision was not guided by a wish to influence economics; Econometricajust happened to be where the best papers on decision making had beenpublished in the past, and we were aspiring to be in that company. In thischoice as in many others, we were lucky. Prospect theory turned out to bethe most significant work we ever did, and our article is among the mostoften cited in the social sciences. Two years later, we published inScience an account of framing effects: the large changes of preferencesthat are sometimes caused by inconsequential variations in the wording ofa choice problem.

During the first five years we spent looking at how people makedecisions, we established a dozen facts about choices between riskyoptions. Several of these facts were in flat contradiction to expected utilitytheory. Some had been observed before, a few were new. Then weconstructed a theory that modified expected utility theory just enough toexplain our collection of observations. That was prospect theory.

Our approach to the problem was in the spirit of a field of psychologycalled psychophysics, which was founded and named by the Germanpsychologist and mystic Gustav Fechner (1801–1887). Fechner wasobsessed with the relation of mind and matter. On one side there is aphysical quantity that can vary, such as the energy of a light, the frequencyof a tone, or an amount of money. On the other side there is a subjectiveexperience of brightness, pitch, or value. Mysteriously, variations of thephysical quantity cause variations in the intensity or quality of the subjectiveexperience. Fechner’s project was to find the psychophysical laws that

(PDF) Thinking, fast and slow by daniel kahneman - PDFSLIDE.US (264)

relate the subjective quantity in the observer’s mind to the objectivequantity in the material world. He proposed that for many dimensions, thefunction is logarithmic—which simply means that an increase of stimulusintensity by a given factor (say, times 1.5 or times 10) always yields thesame increment on the psychological scale. If raising the energy of thesound from 10 to 100 units of physical energy increases psychologicalintensity by 4 units, then a further increase of stimulus intensity from 100 to1,000 will also increase psychological intensity by 4 units.

Bernoulli’s Error

As Fechner well knew, he was not the first to look for a function that relBinepitze="4">utility) and the actual amount of money. He argued that agift of 10 ducats has the same utility to someone who already has 100ducats as a gift of 20 ducats to someone whose current wealth is 200ducats. Bernoulli was right, of course: we normally speak of changes ofincome in terms of percentages, as when we say “she got a 30% raise.”The idea is that a 30% raise may evoke a fairly similar psychologicalresponse for the rich and for the poor, which an increase of $100 will notdo. As in Fechner’s law, the psychological response to a change of wealthis inversely proportional to the initial amount of wealth, leading to theconclusion that utility is a logarithmic function of wealth. If this function isaccurate, the same psychological distance separates $100,000 from $1million, and $10 million from $100 million.

Bernoulli drew on his psychological insight into the utility of wealth topropose a radically new approach to the evaluation of gambles, animportant topic for the mathematicians of his day. Prior to Bernoulli,mathematicians had assumed that gambles are assessed by theirexpected value: a weighted average of the possible outcomes, whereeach outcome is weighted by its probability. For example, the expectedvalue of:

80% chance to win $100 and 20% chance to win $10 is $82 (0.8× 100 + 0.2 × 10).

Now ask yourself this question: Which would you prefer to receive as a gift,this gamble or $80 for sure? Almost everyone prefers the sure thing. Ifpeople valued uncertain prospects by their expected value, they wouldprefer the gamble, because $82 is more than $80. Bernoulli pointed outthat people do not in fact evaluate gambles in this way.

Bernoulli observed that most people dislike risk (the chance of receivingthe lowest possible outcome), and if they are offered a choice between a

(PDF) Thinking, fast and slow by daniel kahneman - PDFSLIDE.US (265)

the lowest possible outcome), and if they are offered a choice between agamble and an amount equal to its expected value they will pick the surething. In fact a risk-averse decision maker will choose a sure thing that isless than expected value, in effect paying a premium to avoid theuncertainty. One hundred years before Fechner, Bernoulli inventedpsychophysics to explain this aversion to risk. His idea wasstraightforward: people’s choices are based not on dollar values but on thepsychological values of outcomes, their utilities. The psychological value ofa gamble is therefore not the weighted average of its possible dollaroutcomes; it is the average of the utilities of these outcomes, eachweighted by its probability.

Table 3 shows a version of the utility function that Bernoulli calculated; itpresents the utility of different levels of wealth, from 1 million to 10 million.You can see that adding 1 million to a wealth of 1 million yields anincrement of 20 utility points, but adding 1 million to a wealth of 9 millionadds only 4 points. Bernoulli proposed that the diminishing marginal valueof wealth (in the modern jargon) is what explains risk aversion—thecommon preference that people generally show for a sure thing over afavorable gamble of equal or slightly higher expected value. Consider thischoice:

Table 3

The expected value of the gamble and the “sure thing” are equal in ducats(4 million), but the psychological utilities of the two options are different,because of the diminishing utility of wealth: the increment of utility from 1million to 4 million is 50 units, but an equal increment, from 4 to 7 million,increases the utility of wealth by only 24 units. The utility of the gamble is94/2 = 47 (the utility of its two outcomes, each weighted by its probability of1/2). The utility of 4 million is 60. Because 60 is more than 47, an individualwith this utility function will prefer the sure thing. Bernoulli’s insight was thata decision maker with diminishing marginal utility for wealth will be riskaverse.

Bernoulli’s essay is a marvel of concise brilliance. He applied his new

(PDF) Thinking, fast and slow by daniel kahneman - PDFSLIDE.US (266)

concept of expected utility (which he called “moral expectation”) tocompute how much a merchant in St. Petersburg would be willing to pay toinsure a shipment of spice from Amsterdam if “he is well aware of the factthat at this time of year of one hundred ships which sail from Amsterdam toPetersburg, five are usually lost.” His utility function explained why poorpeople buy insurance and why richer people sell it to them. As you can seein the table, the loss of 1 million causes a loss of 4 points of utility (from100 to 96) to someone who has 10 million and a much larger loss of 18points (from 48 to 30) to someone who starts off with 3 million. The poorerman will happily pay a premium to transfer the risk to the richer one, whichis what insurance is about. Bernoulli also offered a solution to the famous“St. Petersburg paradox,” in which people who are offered a gamble thathas infinite expected value (in ducats) are willing to spend only a fewducats for it. Most impressive, his analysis of risk attitudes in terms ofpreferences for wealth has stood the test of time: it is still current ineconomic analysis almost 300 years later.

The longevity of the theory is all the more remarkable because it isseriously flawed. The errors of a theory are rarely found in what it assertsexplicitly; they hide in what it ignores or tacitly assumes. For an example,take the following scenarios:

Today Jack and Jill each have a wealth of 5 million.Yesterday, Jack had 1 million and Jill had 9 million.Are they equally happy? (Do they have the same utility?)

Bernoulli’s theory assumes that the utility of their wealth is what makespeople more or less happy. Jack and Jill have the same wealth, and thetheory therefore asserts that they should be equally happy, but you do notneed a degree in psychology to know that today Jack is elated and Jilldespondent. Indeed, we know that Jack would be a great deal happierthan Jill even if he had only 2 million today while she has 5. So Bernoulli’stheory must be wrong.

The happiness that Jack and Jill experience is determined by the recentchange in their wealth, relative to the different states of wealth that definetheir reference points (1 million for Jack, 9 million for Jill). This referencedependence is ubiquitous in sensation and perception. The same soundwill be experienced as very loud or quite faint, depending on whether it waspreceded by a whisper or by a roar. To predict the subjective experienceof loudness, it is not enough to know its absolute energy; you also need toBineli&r quite fa know the reference sound to which it is automaticallycompared. Similarly, you need to know about the background before youcan predict whether a gray patch on a page will appear dark or light. And

(PDF) Thinking, fast and slow by daniel kahneman - PDFSLIDE.US (267)

you need to know the reference before you can predict the utility of anamount of wealth.

For another example of what Bernoulli’s theory misses, considerAnthony and Betty:

Anthony’s current wealth is 1 million.Betty’s current wealth is 4 million.

They are both offered a choice between a gamble and a sure thing.

The gamble: equal chances to end up owning 1 million or 4millionORThe sure thing: own 2 million for sure

In Bernoulli’s account, Anthony and Betty face the same choice: theirexpected wealth will be 2.5 million if they take the gamble and 2 million ifthey prefer the sure-thing option. Bernoulli would therefore expect Anthonyand Betty to make the same choice, but this prediction is incorrect. Hereagain, the theory fails because it does not allow for the different referencepoints from which Anthony and Betty consider their options. If you imagineyourself in Anthony’s and Betty’s shoes, you will quickly see that currentwealth matters a great deal. Here is how they may think:

Anthony (who currently owns 1 million): “If I choose the sure thing,my wealth will double with certainty. This is very attractive.Alternatively, I can take a gamble with equal chances toquadruple my wealth or to gain nothing.”

Betty (who currently owns 4 million): “If I choose the sure thing, Ilose half of my wealth with certainty, which is awful. Alternatively, Ican take a gamble with equal chances to lose three-quarters ofmy wealth or to lose nothing.”

You can sense that Anthony and Betty are likely to make differentchoices because the sure-thing option of owning 2 million makes Anthonyhappy and makes Betty miserable. Note also how the sure outcome differsfrom the worst outcome of the gamble: for Anthony, it is the differencebetween doubling his wealth and gaining nothing; for Betty, it is thedifference between losing half her wealth and losing three-quarters of it.Betty is much more likely to take her chances, as others do when faced

(PDF) Thinking, fast and slow by daniel kahneman - PDFSLIDE.US (268)

with very bad options. As I have told their story, neither Anthony nor Bettythinks in terms of states of wealth: Anthony thinks of gains and Betty thinksof losses. The psychological outcomes they assess are entirely different,although the possible states of wealth they face are the same.

Because Bernoulli’s model lacks the idea of a reference point, expectedutility theory does not represent the obvious fact that the outcome that isgood for Anthony is bad for Betty. His model could explain Anthony’s riskaversion, but it cannot explain Betty’s risk-seeking preference for thegamble, a behavior that is often observed in entrepreneurs and in generalswhen all their options are bad.

All this is rather obvious, isn’t it? One could easily imagine Bernoullihimself constructing similar examples and developing a more complextheory to accommodate them; for some reason, he did not. One could alsoimagine colleagues of his time disagreeing with him, or later scholarsobjecting as they read his essay; for some reason, they did not either.

The mystery is how a conception of the utility of outcomes that isvulnerable to such obvious counterexamples survived for so long. I canexplain it only by a weakness of the scholarly mind that I have oftenobserved in myself. I call it theory-induced blindness: once you haveaccepted a theory and used it as a tool in your thinking, it is extraordinarilydifficult to notice its flaws. If you come upon an observation that does notseem to fit the model, you assume that there must be a perfectly goodexplanation that you are somehow missing. You give the theory the benefitof the doubt, trusting the community of experts who have accepted it. Manyscholars have surely thought at one time or another of stories such asthose of Anthony and Betty, or Jack and Jill, and casually noted that thesestories did not jibe with utility theory. But they did not pursue the idea to thepoint of saying, “This theory is seriously wrong because it ignores the factthat utility depends on the history of one’s wealth, not only on presentwealth.” As the psychologist Daniel Gilbert observed, disbelieving is hardwork, and System 2 is easily tired.

Speaking of Bernoulli’s Errors

“He was very happy with a $20,000 bonus three years ago, buthis salary has gone up by 20% since, so he will need a higherbonus to get the same utility.”

“Both candidates are willing to accept the salary we’re offering,

(PDF) Thinking, fast and slow by daniel kahneman - PDFSLIDE.US (269)

but they won’t be equally satisfied because their reference pointsare different. She currently has a much higher salary.”

“She’s suing him for alimony. She would actually like to settle, buthe prefers to go to court. That’s not surprising—she can onlygain, so she’s risk averse. He, on the other hand, faces optionsthat are all bad, so he’d rather take the risk.”

(PDF) Thinking, fast and slow by daniel kahneman - PDFSLIDE.US (270)

Prospect Theory

Amos and I stumbled on the central flaw in Bernoulli’s theory by a luckycombination of skill and ignorance. At Amos’s suggestion, I read a chapterin his book that described experiments in which distinguished scholarshad measured the utility of money by asking people to make choices aboutgambles in which the participant could win or lose a few pennies. Theexperimenters were measuring the utility of wealth, by modifying wealthwithin a range of less than a dollar. This raised questions. Is it plausible toassume that people evaluate the gambles by tiny differences in wealth?How could one hope to learn about the psychophysics of wealth bystudying reactions to gains and losses of pennies? Recent developmentsin psychophysical theory suggested that if you want to study the subjectivevalue of wealth, you shou Clth"ld ask direct questions about wealth, notabout changes of wealth. I did not know enough about utility theory to beblinded by respect for it, and I was puzzled.

When Amos and I met the next day, I reported my difficulties as a vaguethought, not as a discovery. I fully expected him to set me straight and toexplain why the experiment that had puzzled me made sense after all, buthe did nothing of the kind—the relevance of the modern psychophysicswas immediately obvious to him. He remembered that the economist HarryMarkowitz, who would later earn the Nobel Prize for his work on finance,had proposed a theory in which utilities were attached to changes ofwealth rather than to states of wealth. Markowitz’s idea had been aroundfor a quarter of a century and had not attracted much attention, but wequickly concluded that this was the way to go, and that the theory we wereplanning to develop would define outcomes as gains and losses, not asstates of wealth. Knowledge of perception and ignorance about decisiontheory both contributed to a large step forward in our research.

We soon knew that we had overcome a serious case of theory-inducedblindness, because the idea we had rejected now seemed not only falsebut absurd. We were amused to realize that we were unable to assess ourcurrent wealth within tens of thousands of dollars. The idea of derivingattitudes to small changes from the utility of wealth now seemedindefensible. You know you have made a theoretical advance when youcan no longer reconstruct why you failed for so long to see the obvious.Still, it took us years to explore the implications of thinking about outcomesas gains and losses.

In utility theory, the utility of a gain is assessed by comparing the utilitiesof two states of wealth. For example, the utility of getting an extra $500when your wealth is $1 million is the difference between the utility of

(PDF) Thinking, fast and slow by daniel kahneman - PDFSLIDE.US (271)

$1,000,500 and the utility of $1 million. And if you own the larger amount,the disutility of losing $500 is again the difference between the utilities ofthe two states of wealth. In this theory, the utilities of gains and losses areallowed to differ only in their sign (+ or –). There is no way to represent thefact that the disutility of losing $500 could be greater than the utility ofwinning the same amount—though of course it is. As might be expected ina situation of theory-induced blindness, possible differences betweengains and losses were neither expected nor studied. The distinctionbetween gains and losses was assumed not to matter, so there was nopoint in examining it.

Amos and I did not see immediately that our focus on changes of wealthopened the way to an exploration of a new topic. We were mainlyconcerned with differences between gambles with high or low probabilityof winning. One day, Amos made the casual suggestion, “How aboutlosses?” and we quickly found that our familiar risk aversion was replacedby risk seeking when we switched our focus. Consider these twoproblems:

Problem 1: Which do you choose?Get $900 for sure OR 90% chance to get $1,000

Problem 2: Which do you choose?Lose $900 for sure OR 90% chance to lose $1,000

You were probably risk averse in problem 1, as is the great majority ofpeople. The subjective value of a gain of $900 is certainly more than 90%of the value of a ga Blth"it ue of a gin of $1,000. The risk-averse choice inthis problem would not have surprised Bernoulli.

Now examine your preference in problem 2. If you are like most otherpeople, you chose the gamble in this question. The explanation for thisrisk-seeking choice is the mirror image of the explanation of risk aversionin problem 1: the (negative) value of losing $900 is much more than 90% ofthe (negative) value of losing $1,000. The sure loss is very aversive, andthis drives you to take the risk. Later, we will see that the evaluations of theprobabilities (90% versus 100%) also contributes to both risk aversion inproblem 1 and the preference for the gamble in problem 2.

We were not the first to notice that people become risk seeking when alltheir options are bad, but theory-induced blindness had prevailed.Because the dominant theory did not provide a plausible way toaccommodate different attitudes to risk for gains and losses, the fact thatthe attitudes differed had to be ignored. In contrast, our decision to view

(PDF) Thinking, fast and slow by daniel kahneman - PDFSLIDE.US (272)

outcomes as gains and losses led us to focus precisely on thisdiscrepancy. The observation of contrasting attitudes to risk with favorableand unfavorable prospects soon yielded a significant advance: we found away to demonstrate the central error in Bernoulli’s model of choice. Have alook:

Problem 3: In addition to whatever you own, you have been given$1,000.You are now asked to choose one of these options:50% chance to win $1,000 OR get $500 for sure

Problem 4: In addition to whatever you own, you have been given$2,000.You are now asked to choose one of these options:50% chance to lose $1,000 OR lose $500 for sure

You can easily confirm that in terms of final states of wealth—all thatmatters for Bernoulli’s theory—problems 3 and 4 are identical. In bothcases you have a choice between the same two options: you can have thecertainty of being richer than you currently are by $1,500, or accept agamble in which you have equal chances to be richer by $1,000 or by$2,000. In Bernoulli’s theory, therefore, the two problems should elicitsimilar preferences. Check your intuitions, and you will probably guesswhat other people did.

In the first choice, a large majority of respondents preferred the surething.In the second choice, a large majority preferred the gamble.

The finding of different preferences in problems 3 and 4 was a decisivecounterexample to the key idea of Bernoulli’s theory. If the utility of wealth isall that matters, then transparently equivalent statements of the sameproblem should yield identical choices. The comparison of the problemshighlights the all-important role of the reference point from which theoptions are evaluated. The reference point is higher than current wealth by$1,000 in problem 3, by $2,000 in problem 4. Being richer by $1,500 istherefore a gain of $500 in problem 3 and a loss in problem 4. Obviously,other examples of the same kind are easy to generate. The story of

(PDF) Thinking, fast and slow by daniel kahneman - PDFSLIDE.US (273)

Anthony and Betty had a similar structure.How much attention did you pay to the gift of $1,000 or $2,000 that

you were “given” prior to making your choice? If you are like most people,you barely noticed it. Indeed, there was no reason for you to attend to it,because the gift is included in the reference point, and reference pointsare generally ignored. You know something about your preferences thatutility theorists do not—that your attitudes to risk would not be different ifyour net worth were higher or lower by a few thousand dollars (unless youare abjectly poor). And you also know that your attitudes to gains andlosses are not derived from your evaluation of your wealth. The reason youlike the idea of gaining $100 and dislike the idea of losing $100 is not thatthese amounts change your wealth. You just like winning and dislike losing—and you almost certainly dislike losing more than you like winning.

The four problems highlight the weakness of Bernoulli’s model. Histheory is too simple and lacks a moving part. The missing variable is thereference point, the earlier state relative to which gains and losses areevaluated. In Bernoulli’s theory you need to know only the state of wealth todetermine its utility, but in prospect theory you also need to know thereference state. Prospect theory is therefore more complex than utilitytheory. In science complexity is considered a cost, which must be justifiedby a sufficiently rich set of new and (preferably) interesting predictions offacts that the existing theory cannot explain. This was the challenge we hadto meet.

Although Amos and I were not working with the two-systems model ofthe mind, it’s clear now that there are three cognitive features at the heartof prospect theory. They play an essential role in the evaluation of financialoutcomes and are common to many automatic processes of perception,judgment, and emotion. They should be seen as operating characteristicsof System 1.

Evaluation is relative to a neutral reference point, which issometimes referred to as an “adaptation level.” You can easily set upa compelling demonstration of this principle. Place three bowls ofwater in front of you. Put ice water into the left-hand bowl and warmwater into the right-hand bowl. The water in the middle bowl shouldbe at room temperature. Immerse your hands in the cold and warmwater for about a minute, then dip both in the middle bowl. You willexperience the same temperature as heat in one hand and cold inthe other. For financial outcomes, the usual reference point is thestatus quo, but it can also be the outcome that you expect, or

(PDF) Thinking, fast and slow by daniel kahneman - PDFSLIDE.US (274)

perhaps the outcome to which you feel entitled, for example, theraise or bonus that your colleagues receive. Outcomes that arebetter than the reference points are gains. Below the reference pointthey are losses.A principle of diminishing sensitivity applies to both sensorydimensions and the evaluation of changes of wealth. Turning on aweak light has a large effect in a dark room. The same increment oflight may be undetectable in a brightly illuminated room. Similarly, thesubjective difference between $900 and $1,000 is much smaller thanthe difference between $100 and $200.The third principle is loss aversion. When directly compared orweighted against each other, losses loom larger than gains. Thisasymmetry between the power of positive and negative expectationsor experiences has an evolutionary history. Organisms that treatthreats as more urgent than opportunities have a better chance tosurvive and reproduce.

The three principles that govern the value of outcomes are illustrated byfigure 1 Blth" wagure 0. If prospect theory had a flag, this image would bedrawn on it. The graph shows the psychological value of gains and losses,which are the “carriers” of value in prospect theory (unlike Bernoulli’smodel, in which states of wealth are the carriers of value). The graph hastwo distinct parts, to the right and to the left of a neutral reference point. Asalient feature is that it is S-shaped, which represents diminishingsensitivity for both gains and losses. Finally, the two curves of the S are notsymmetrical. The slope of the function changes abruptly at the referencepoint: the response to losses is stronger than the response tocorresponding gains. This is loss aversion.

(PDF) Thinking, fast and slow by daniel kahneman - PDFSLIDE.US (275)

Figure 10

Loss Aversion

Many of the options we face in life are “mixed”: there is a risk of loss andan opportunity for gain, and we must decide whether to accept the gambleor reject it. Investors who evaluate a start-up, lawyers who wonder whetherto file a lawsuit, wartime generals who consider an offensive, andpoliticians who must decide whether to run for office all face thepossibilities of victory or defeat. For an elementary example of a mixedprospect, examine your reaction to the next question.

Problem 5: You are offered a gamble on the toss of a coin.If the coin shows tails, you lose $100.If the coin shows heads, you win $150.Is this gamble attractive? Would you accept it?

To make this choice, you must balance the psychological benefit of getting$150 against the psychological cost of losing $100. How do you feel aboutit? Although the expected value of the gamble is obviously positive,

(PDF) Thinking, fast and slow by daniel kahneman - PDFSLIDE.US (276)

because you stand to gain more than you can lose, you probably dislike it—most people do. The rejection of this gamble is an act of System 2, butthe critical inputs are emotional responses that are generated by System1. For most people, the fear of losing $100 is more intense than the hopeof gaining $150. We concluded from many such observations that “lossesloom larger than gains” and that people are loss averse.

You can measure the extent of your aversion to losses by asking yourselfa question: What is the smallest gain that I need to balance an equalchance to lose $100? For many people the answer is about $200, twice asmuch as the loss. The “loss aversion ratio” has been estimated in severalexperiments and is usually in the range of 1.5 to 2.5. This is an average, ofcourse; some people are much more loss averse than others. Professionalrisk takers in the financial markets are more tolerant of losses, probablybecause they do not respond emotionally to every fluctuation. Whenparticipants in an experiment were instructed to “think like a trader,” theybecame less loss averse and their emotional reaction to losses (measuredby a physiological index of emotional arousal) was sharply reduced.

In order to examine your loss aversion ratio for different stakes, considerthe following questions. Ignore any social considerations, do not try toappear either bold Blth"vioher or cautious, and focus only on the subjectiveimpact of the possible loss and the off setting gain.

Consider a 5 0–5 0 gamble in which you can lose $10. What is thesmallest gain that makes the gamble attractive? If you say $10, thenyou are indifferent to risk. If you give a number less than $10, youseek risk. If your answer is above $10, you are loss averse.What about a possible loss of $500 on a coin toss? What possiblegain do you require to off set it?What about a loss of $2,000?

As you carried out this exercise, you probably found that your loss aversioncoefficient tends to increase when the stakes rise, but not dramatically. Allbets are off, of course, if the possible loss is potentially ruinous, or if yourlifestyle is threatened. The loss aversion coefficient is very large in suchcases and may even be infinite—there are risks that you will not accept,regardless of how many millions you might stand to win if you are lucky.

Another look at figure 10 may help prevent a common confusion. In thischapter I have made two claims, which some readers may view ascontradictory:

(PDF) Thinking, fast and slow by daniel kahneman - PDFSLIDE.US (277)

In mixed gambles, where both a gain and a loss are possible, lossaversion causes extremely risk-averse choices.In bad choices, where a sure loss is compared to a larger loss that ismerely probable, diminishing sensitivity causes risk seeking.

There is no contradiction. In the mixed case, the possible loss looms twiceas large as the possible gain, as you can see by comparing the slopes ofthe value function for losses and gains. In the bad case, the bending of thevalue curve (diminishing sensitivity) causes risk seeking. The pain of losing$900 is more than 90% of the pain of losing $1,000. These two insightsare the essence of prospect theory.

Figure 10 shows an abrupt change in the slope of the value function wheregains turn into losses, because there is considerable loss aversion evenwhen the amount at risk is minuscule relative to your wealth. Is it plausiblethat attitudes to states of wealth could explain the extreme aversion tosmall risks? It is a striking example of theory-induced blindness that thisobvious flaw in Bernoulli’s theory failed to attract scholarly notice for morethan 250 years. In 2000, the behavioral economist Matthew Rabin finallyproved mathematically that attempts to explain loss aversion by the utility ofwealth are absurd and doomed to fail, and his proof attracted attention.Rabin’s theorem shows that anyone who rejects a favorable gamble withsmall stakes is mathematically committed to a foolish level of risk aversionfor some larger gamble. For example, he notes that most Humans rejectthe following gamble:

50% chance to lose $100 and 50% chance to win $200

He then shows that according to utility theory, an individual who rejects thatgamble will also turn down the following gamble:

50% chance to lose $200 and 50% chance to win $20,000

But of course no one in his or her right mind will reject this gamble! In anexuberant article they wrote abo Blth"ins>

Perhaps carried away by their enthusiasm, they concluded their articleby recalling the famous Monty Python sketch in which a frustrated customer

(PDF) Thinking, fast and slow by daniel kahneman - PDFSLIDE.US (278)

attempts to return a dead parrot to a pet store. The customer uses a longseries of phrases to describe the state of the bird, culminating in “this is anex-parrot.” Rabin and Thaler went on to say that “it is time for economiststo recognize that expected utility is an ex-hypothesis.” Many economistssaw this flippant statement as little short of blasphemy. However, thetheory-induced blindness of accepting the utility of wealth as anexplanation of attitudes to small losses is a legitimate target for humorouscomment.

Blind Spots pf Prospect Theory

So far in this part of the book I have extolled the virtues of prospect theoryand criticized the rational model and expected utility theory. It is time forsome balance.

Most graduate students in economics have heard about prospect theoryand loss aversion, but you are unlikely to find these terms in the index of anintroductory text in economics. I am sometimes pained by this omission,but in fact it is quite reasonable, because of the central role of rationality inbasic economic theory. The standard concepts and results thatundergraduates are taught are most easily explained by assuming thatEcons do not make foolish mistakes. This assumption is truly necessary,and it would be undermined by introducing the Humans of prospect theory,whose evaluations of outcomes are unreasonably short-sighted.

There are good reasons for keeping prospect theory out of introductorytexts. The basic concepts of economics are essential intellectual tools,which are not easy to grasp even with simplified and unrealisticassumptions about the nature of the economic agents who interact inmarkets. Raising questions about these assumptions even as they areintroduced would be confusing, and perhaps demoralizing. It is reasonableto put priority on helping students acquire the basic tools of the discipline.Furthermore, the failure of rationality that is built into prospect theory isoften irrelevant to the predictions of economic theory, which work out withgreat precision in some situations and provide good approximations inmany others. In some contexts, however, the difference becomessignificant: the Humans described by prospect theory are guided by theimmediate emotional impact of gains and losses, not by long-termprospects of wealth and global utility.

I emphasized theory-induced blindness in my discussion of flaws inBernoulli’s model that remained unquestioned for more than two centuries.But of course theory-induced blindness is not restricted to expected utilitytheory. Prospect theory has flaws of its own, and theory-induced blindness

(PDF) Thinking, fast and slow by daniel kahneman - PDFSLIDE.US (279)

to these flaws has contributed to its acceptance as the main alternative toutility theory.

Consider the assumption of prospect theory, that the reference point,usually the status quo, has a value of zero. This assumption seemsreasonable, but it leads to some absurd consequences. Have a good lookat the following prospects. What would it be like to own them?

A. one chance in a million to win $1 millionB. 90% chance to win $12 and 10% chance to win nothingC. 90% chance to win $1 million and 10% chance to win nothing

Winning nothing is a possible outcome in all three gambles, and prospecttheory assigns the same value to that outcome in the three cases. Winningnothing is the reference point and its value is zero. Do these statementscorrespond to your experience? Of course not. Winning nothing is anonevent in the first two cases, and assigning it a value of zero makesgood sense. In contrast, failing to win in the third scenario is intenselydisappointing. Like a salary increase that has been promised informally,the high probability of winning the large sum sets up a tentative newreference point. Relative to your expectations, winning nothing will beexperienced as a large loss. Prospect theory cannot cope with this fact,because it does not allow the value of an outcome (in this case, winningnothing) to change when it is highly unlikely, or when the alternative is veryvaluable. In simple words, prospect theory cannot deal withdisappointment. Disappointment and the anticipation of disappointmentare real, however, and the failure to acknowledge them is as obvious aflow as the counterexamples that I invoked to criticize Bernoulli’s theory.

Prospect theory and utility theory also fail to allow for regret. The twotheories share the assumption that available options in a choice areevaluated separately and independently, and that the option with thehighest value is selected. This assumption is certainly wrong, as thefollowing example shows.

Problem 6: Choose between 90% chance to win $1 million OR$50 with certainty.

Problem 7: Choose between 90% chance to win $1 million OR$150,000 with certainty.

Compare the anticipated pain of choosing the gamble and not winning inthe two cases. Failing to win is a disappointment in both, but the potential

(PDF) Thinking, fast and slow by daniel kahneman - PDFSLIDE.US (280)

pain is compounded in problem 7 by knowing that if you choose thegamble and lose you will regret the “greedy” decision you made byspurning a sure gift of $150,000. In regret, the experience of an outcomedepends on an option you could have adopted but did not.

Several economists and psychologists have proposed models ofdecision making that are based on the emotions of regret anddisappointment. It is fair to say that these models have had less influencethan prospect theory, and the reason is instructive. The emotions of regretand disappointment are real, and decision makers surely anticipate theseemotions when making their choices. The problem is that regret theoriesmake few striking predictions that would distinguish them from prospecttheory, which has the advantage of being simpler. The complexity ofprospect theory was more acceptable in the competition with expectedutility theory because it did predict observations that expected utility theorycould not explain.

Richer and more realistic assumptions do not suffice to make a theorysuccessful. Scientists use theories as a bag of working tools, and they willnot take on the burden of a heavier bag unless the new tools are veryuseful. Prospect theory was accepted by many scholars not because it is“true” but because the concepts that it added to utility theory, notably thereference point and loss aversion, were worth the trouble; they yielded newpredictions that turned out to be true. We were lucky.

Speaking of Prospect Theory

“He suffers from extreme loss aversion, which makes him turn down veryfavorable opportunities.”

“Considering her vast wealth, her emotional response to trivial gains andlosses makes no sense.”

“He weighs losses about twice as much as gains, which is normal.”

(PDF) Thinking, fast and slow by daniel kahneman - PDFSLIDE.US (281)

The Endowment Effect

You have probably seen figure 11 or a close cousin of it even if you neverhad a class in economics. The graph displays an individual’s “indifferencemap” for two goods.

Figure 11

Students learn in introductory economics classes that each point on themap specifies a particular combination of income and vacation days. Each“indifference curve” connects the combinations of the two goods that areequally desirable—they have the same utility. The curves would turn intoparallel straight lines if people were willing to “sell” vacation days for extraincome at the same price regardless of how much income and how muchvacation time they have. The convex shape indicates diminishing marginalutility: the more leisure you have, the less you care for an extra day of it,and each added day is worth less than the one before. Similarly, the moreincome you have, the less you care for an extra dollar, and the amount youare willing to give up for an extra day of leisure increases.

All locations on an indifference curve are equally attractive. This isliterally what indifference means: you don’t care where you are on an

(PDF) Thinking, fast and slow by daniel kahneman - PDFSLIDE.US (282)

indifference curve. So if A and B are on the same indifference curve foryou, you are indifferent between them and will need no incentive to movefrom one to the other, or back. Some version of this figure has appeared inevery economics textbook written in the last hundred years, and manymillions of students have stared at it. Few have noticed what is missing.Here again, the power and elegance of a theoretical model have blindedstudents and scholars to a serious deficiency.

What is missing from the figure is an indication of the individual’s currentincome and leisure. If you are a salaried employee, the terms of youremployment specify a salary and a number of vacation days, which is apoint on the map. This is your reference point, your status quo, but thefigure does not show it. By failing to display it, the theorists who draw thisfigure invite you to believe that the reference point does not matter, but bynow you know that of course it does. This is Bernoulli’s error all over again.The representation of indifference curves implicitly assumes that your utilityat any given moment is determined entirely by your present situation, thatthe past is irrelevant, and that your evaluation of a possible job does notdepend on the terms of your current job. These assumptions arecompletely unrealistic in this case and in many others.

The omission of the ref Con serence point from the indifference map is asurprising case of theory-induced blindness, because we so oftenencounter cases in which the reference point obviously matters. In labornegotiations, it is well understood by both sides that the reference point isthe existing contract and that the negotiations will focus on mutualdemands for concessions relative to that reference point. The role of lossaversion in bargaining is also well understood: making concessions hurts.You have much personal experience of the role of reference point. If youchanged jobs or locations, or even considered such a change, you surelyremember that the features of the new place were coded as pluses orminuses relative to where you were. You may also have noticed thatdisadvantages loomed larger than advantages in this evaluation—lossaversion was at work. It is difficult to accept changes for the worse. Forexample, the minimal wage that unemployed workers would accept for newemployment averages 90% of their previous wage, and it drops by lessthan 10% over a period of one year.

To appreciate the power that the reference point exerts on choices,consider Albert and Ben, “hedonic twins” who have identical tastes andcurrently hold identical starting jobs, with little income and little leisure time.Their current circumstances correspond to the point marked 1 in figure 11.The firm offers them two improved positions, A and B, and lets themdecide who will get a raise of $10,000 (position A) and who will get anextra day of paid vacation each month (position B). As they are both

(PDF) Thinking, fast and slow by daniel kahneman - PDFSLIDE.US (283)

indifferent, they toss a coin. Albert gets the raise, Ben gets the extraleisure. Some time passes as the twins get accustomed to their positions.Now the company suggests they may switch jobs if they wish.

The standard theory represented in the figure assumes that preferencesare stable over time. Positions A and B are equally attractive for both twinsand they will need little or no incentive to switch. In sharp contrast, prospecttheory asserts that both twins will definitely prefer to remain as they are.This preference for the status quo is a consequence of loss aversion.

Let us focus on Albert. He was initially in position 1 on the graph, andfrom that reference point he found these two alternatives equally attractive:

Go to A: a raise of $10,000ORGo to B: 12 extra days of vacation

Taking position A changes Albert’s reference point, and when heconsiders switching to B, his choice has a new structure:

Stay at A: no gain and no lossORMove to B: 12 extra days of vacation and a $10,000 salary cut

You just had the subjective experience of loss aversion. You could feel it: asalary cut of $10,000 is very bad news. Even if a gain of 12 vacation dayswas as impressive as a gain of $10,000, the same improvement of leisureis not sufficient to compensate for a loss of $10,000. Albert will stay at Abecause the disadvantage of moving outweighs the advantage. The samereasoning applies to Ben, who will also want to keep his present jobbecause the loss of now-precious leisure outweighs the benefit of the extraincome.

This example highlights two aspects of choice that the st Bon s Ae stBonandard model of indifference curves does not predict. First, tastes arenot fixed; they vary with the reference point. Second, the disadvantages ofa change loom larger than its advantages, inducing a bias that favors thestatus quo. Of course, loss aversion does not imply that you never prefer tochange your situation; the benefits of an opportunity may exceed evenoverweighted losses. Loss aversion implies only that choices are stronglybiased in favor of the reference situation (and generally biased to favorsmall rather than large changes).

Conventional indifference maps and Bernoulli’s representation ofoutcomes as states of wealth share a mistaken assumption: that your utilityfor a state of affairs depends only on that state and is not affected by your

(PDF) Thinking, fast and slow by daniel kahneman - PDFSLIDE.US (284)

history. Correcting that mistake has been one of the achievements ofbehavioral economics.

The Endowment Effect

The question of when an approach or a movement got its start is oftendifficult to answer, but the origin of what is now known as behavioraleconomics can be specified precisely. In the early 1970s, Richard Thaler,then a graduate student in the very conservative economics department ofthe University of Rochester, began having heretical thoughts. Thaler alwayshad a sharp wit and an ironic bent, and as a student he amused himself bycollecting observations of behavior that the model of rational economicbehavior could not explain. He took special pleasure in evidence ofeconomic irrationality among his professors, and he found one that wasparticularly striking.

Professor R (now revealed to be Richard Rosett, who went on tobecome the dean of the University of Chicago Graduate School ofBusiness) was a firm believer in standard economic theory as well as asophisticated wine lover. Thaler observed that Professor R was veryreluctant to sell a bottle from his collection—even at the high price of $100(in 1975 dollars!). Professor R bought wine at auctions, but would neverpay more than $35 for a bottle of that quality. At prices between $35 and$100, he would neither buy nor sell. The large gap is inconsistent witheconomic theory, in which the professor is expected to have a single valuefor the bottle. If a particular bottle is worth $50 to him, then he should bewilling to sell it for any amount in excess of $50. If he did not own the bottle,he should be willing to pay any amount up to $50 for it. The just-acceptableselling price and the just-acceptable buying price should have beenidentical, but in fact the minimum price to sell ($100) was much higher thanthe maximum buying price of $35. Owning the good appeared to increaseits value.

Richard Thaler found many examples of what he called the endowmenteffect, especially for goods that are not regularly traded. You can easilyimagine yourself in a similar situation. Suppose you hold a ticket to a sold-out concert by a popular band, which you bought at the regular price of$200. You are an avid fan and would have been willing to pay up to $500for the ticket. Now you have your ticket and you learn on the Internet thatricher or more desperate fans are offering $3,000. Would you sell? If youresemble most of the audience at sold-out events you do not sell. Yourlowest selling price is above $3,000 and your maximum buying price is$500. This is an example of an endowment effect, and a believer in

(PDF) Thinking, fast and slow by daniel kahneman - PDFSLIDE.US (285)

standard economic theory would be puzzled by it. Thaler was looking for anaccount that could explain puzzles of this kind.

Chance intervened when Thaler met one of our former students at aconference and obtained an early draft of prospect theory. He reports thathe read the manuscript with considerable Bon s Able Bonexcitement,because he quickly realized that the loss-averse value function of prospecttheory could explain the endowment effect and some other puzzles in hiscollection. The solution was to abandon the standard idea that Professor Rhad a unique utility for the state of having a particular bottle. Prospecttheory suggested that the willingness to buy or sell the bottle depends onthe reference point—whether or not the professor owns the bottle now. If heowns it, he considers the pain of giving up the bottle. If he does not own it,he considers the pleasure of getting the bottle. The values were unequalbecause of loss aversion: giving up a bottle of nice wine is more painfulthan getting an equally good bottle is pleasurable. Remember the graph oflosses and gains in the previous chapter. The slope of the function issteeper in the negative domain; the response to a loss is stronger than theresponse to a corresponding gain. This was the explanation of theendowment effect that Thaler had been searching for. And the firstapplication of prospect theory to an economic puzzle now appears to havebeen a significant milestone in the development of behavioral economics.

Thaler arranged to spend a year at Stanford when he knew that Amosand I would be there. During this productive period, we learned much fromeach other and became friends. Seven years later, he and I had anotheropportunity to spend a year together and to continue the conversationbetween psychology and economics. The Russell Sage Foundation, whichwas for a long time the main sponsor of behavioral economics, gave oneof its first grants to Thaler for the purpose of spending a year with me inVancouver. During that year, we worked closely with a local economist,Jack Knetsch, with whom we shared intense interest in the endowmenteffect, the rules of economic fairness, and spicy Chinese food.

The starting point for our investigation was that the endowment effect isnot universal. If someone asks you to change a $5 bill for five singles, youhand over the five ones without any sense of loss. Nor is there much lossaversion when you shop for shoes. The merchant who gives up the shoesin exchange for money certainly feels no loss. Indeed, the shoes that hehands over have always been, from his point of view, a cumbersome proxyfor money that he was hoping to collect from some consumer. Furthermore,you probably do not experience paying the merchant as a loss, becauseyou were effectively holding money as a proxy for the shoes you intendedto buy. These cases of routine trading are not essentially different from the

(PDF) Thinking, fast and slow by daniel kahneman - PDFSLIDE.US (286)

exchange of a $5 bill for five singles. There is no loss aversion on eitherside of routine commercial exchanges.

What distinguishes these market transactions from Professor R’sreluctance to sell his wine, or the reluctance of Super Bowl ticket holders tosell even at a very high price? The distinctive feature is that both the shoesthe merchant sells you and the money you spend from your budget forshoes are held “for exchange.” They are intended to be traded for othergoods. Other goods, such as wine and Super Bowl tickets, are held “foruse,” to be consumed or otherwise enjoyed. Your leisure time and thestandard of living that your income supports are also not intended for saleor exchange.

Knetsch, Thaler, and I set out to design an experiment that wouldhighlight the contrast between goods that are held for use and forexchange. We borrowed one aspect of the design of our experiment fromVernon Smith, the founder of experimental economics, with whom I wouldshare a Nobel Prize many years later. In this method, a limited number oftokens are distributed to the participants in a “market.” Any participantswho own a token at the end Bon s A end Bon of the experiment canredeem it for cash. The redemption values differ for different individuals, torepresent the fact that the goods traded in markets are more valuable tosome people than to others. The same token may be worth $10 to you and$20 to me, and an exchange at any price between these values will beadvantageous to both of us.

Smith created vivid demonstrations of how well the basic mechanismsof supply and demand work. Individuals would make successive publicoffers to buy or sell a token, and others would respond publicly to the offer.Everyone watches these exchanges and sees the price at which thetokens change hands. The results are as regular as those of ademonstration in physics. As inevitably as water flows downhill, those whoown a token that is of little value to them (because their redemption valuesare low) end up selling their token at a profit to someone who values itmore. When trading ends, the tokens are in the hands of those who can getthe most money for them from the experimenter. The magic of the marketshas worked! Furthermore, economic theory correctly predicts both the finalprice at which the market will settle and the number of tokens that willchange hands. If half the participants in the market were randomlyassigned tokens, the theory predicts that half of the tokens will changehands.

We used a variation on Smith’s method for our experiment. Eachsession began with several rounds of trades for tokens, which perfectlyreplicated Smith’s finding. The estimated number of trades was typicallyvery close or identical to the amount predicted by the standard theory. The

(PDF) Thinking, fast and slow by daniel kahneman - PDFSLIDE.US (287)

tokens, of course, had value only because they could be exchanged for theexperimenter’s cash; they had no value for use. Then we conducted asimilar market for an object that we expected people to value for use: anattractive coffee mug, decorated with the university insignia of wherever wewere conducting the experiments. The mug was then worth about $6 (andwould be worth about double that amount today). Mugs were distributedrandomly to half the participants. The Sellers had their mug in front of them,and the Buyers were invited to look at their neighbor’s mug; all indicatedthe price at which they would trade. The Buyers had to use their ownmoney to acquire a mug. The results were dramatic: the average sellingprice was about double the average buying price, and the estimatednumber of trades was less than half of the number predicted by standardtheory. The magic of the market did not work for a good that the ownersexpected to use.

We conducted a series of experiments using variants of the sameprocedure, always with the same results. My favorite is one in which weadded to the Sellers and Buyers a third group—Choosers. Unlike theBuyers, who had to spend their own money to acquire the good, theChoosers could receive either a mug or a sum of money, and theyindicated the amount of money that was as desirable as receiving thegood. These were the results:

Sellers $7.12Choosers $3.12Buyers $2.87

The gap between Sellers and Choosers is remarkable, because theyactually face the same choice! If you are a Seller you can go home witheither a m Bon s A a m Bonug or money, and if you are a Chooser youhave exactly the same two options. The long-term effects of the decisionare identical for the two groups. The only difference is in the emotion of themoment. The high price that Sellers set reflects the reluctance to give upan object that they already own, a reluctance that can be seen in babieswho hold on fiercely to a toy and show great agitation when it is takenaway. Loss aversion is built into the automatic evaluations of System 1.

Buyers and Choosers set similar cash values, although the Buyers haveto pay for the mug, which is free for the Choosers. This is what we wouldexpect if Buyers do not experience spending money on the mug as a loss.Evidence from brain imaging confirms the difference. Selling goods thatone would normally use activates regions of the brain that are associatedwith disgust and pain. Buying also activates these areas, but only when the

(PDF) Thinking, fast and slow by daniel kahneman - PDFSLIDE.US (288)

prices are perceived as too high—when you feel that a seller is takingmoney that exceeds the exchange value. Brain recordings also indicatethat buying at especially low prices is a pleasurable event.

The cash value that the Sellers set on the mug is a bit more than twiceas high as the value set by Choosers and Buyers. The ratio is very close tothe loss aversion coefficient in risky choice, as we might expect if thesame value function for gains and losses of money is applied to bothriskless and risky decisions. A ratio of about 2:1 has appeared in studiesof diverse economic domains, including the response of households toprice changes. As economists would predict, customers tend to increasetheir purchases of eggs, orange juice, or fish when prices drop and toreduce their purchases when prices rise; however, in contrast to thepredictions of economic theory, the effect of price increases (lossesrelative to the reference price) is about twice as large as the effect ofgains.

The mugs experiment has remained the standard demonstration of theendowment effect, along with an even simpler experiment that JackKnetsch reported at about the same time. Knetsch asked two classes to fillout a questionnaire and rewarded them with a gift that remained in front ofthem for the duration of the experiment. In one session, the prize was anexpensive pen; in another, a bar of Swiss chocolate. At the end of theclass, the experimenter showed the alternative gift and allowed everyoneto trade his or her gift for another. Only about 10% of the participants optedto exchange their gift. Most of those who had received the pen stayed withthe pen, and those who had received the chocolate did not budge either.

Thinking Like a Trader

The fundamental ideas of prospect theory are that reference points exist,and that losses loom larger than corresponding gains. Observations in realmarkets collected over the years illustrate the power of these concepts. Astudy of the market for condo apartments in Boston during a downturnyielded particularly clear results. The authors of that study compared thebehavior of owners of similar units who had bought their dwellings atdifferent prices. For a rational agent, the buying price is irrelevant history—the current market value is all that matters. Not so for Humans in a downmarket for housing. Owners who have a high reference point and thus facehigher losses set a higher price on their dwelling, spend a longer timetrying to sell their home, and eventually receive more money.

The original demonstration of an asymmetry between selling prices andbuying prices (or, more convincingly, between selling and choosing) was

(PDF) Thinking, fast and slow by daniel kahneman - PDFSLIDE.US (289)

very important in the initial acceptance of the ideas of reference point andloss aversi Bon s Aersi Bonon. However, it is well understood thatreference points are labile, especially in unusual laboratory situations, andthat the endowment effect can be eliminated by changing the referencepoint.

No endowment effect is expected when owners view their goods ascarriers of value for future exchanges, a widespread attitude in routinecommerce and in financial markets. The experimental economist JohnList, who has studied trading at baseball card conventions, found thatnovice traders were reluctant to part with the cards they owned, but that thisreluctance eventually disappeared with trading experience. Moresurprisingly, List found a large effect of trading experience on theendowment effect for new goods.

At a convention, List displayed a notice that invited people to take part ina short survey, for which they would be compensated with a small gift: acoffee mug or a chocolate bar of equal value. The gift s were assigned atrandom. As the volunteers were about to leave, List said to each of them,“We gave you a mug [or chocolate bar], but you can trade for a chocolatebar [or mug] instead, if you wish.” In an exact replication of Jack Knetsch’searlier experiment, List found that only 18% of the inexperienced traderswere willing to exchange their gift for the other. In sharp contrast,experienced traders showed no trace of an endowment effect: 48% ofthem traded! At least in a market environment in which trading was thenorm, they showed no reluctance to trade.

Jack Knetsch also conducted experiments in which subtle manipulationsmade the endowment effect disappear. Participants displayed anendowment effect only if they had physical possession of the good for awhile before the possibility of trading it was mentioned. Economists of thestandard persuasion might be tempted to say that Knetsch had spent toomuch time with psychologists, because his experimental manipulationshowed concern for the variables that social psychologists expect to beimportant. Indeed, the different methodological concerns of experimentaleconomists and psychologists have been much in evidence in the ongoingdebate about the endowment effect.

Veteran traders have apparently learned to ask the correct question,which is “How much do I want to have that mug, compared with otherthings I could have instead?” This is the question that Econs ask, and withthis question there is no endowment effect, because the asymmetrybetween the pleasure of getting and the pain of giving up is irrelevant.

Recent studies of the psychology of “decision making under poverty”suggest that the poor are another group in which we do not expect to findthe endowment effect. Being poor, in prospect theory, is living below one’s

(PDF) Thinking, fast and slow by daniel kahneman - PDFSLIDE.US (290)

the endowment effect. Being poor, in prospect theory, is living below one’sreference point. There are goods that the poor need and cannot afford, sothey are always “in the losses.” Small amounts of money that they receiveare therefore perceived as a reduced loss, not as a gain. The money helpsone climb a little toward the reference point, but the poor always remain onthe steep limb of the value function.

People who are poor think like traders, but the dynamics are quitedifferent. Unlike traders, the poor are not indifferent to the differencesbetween gaining and giving up. Their problem is that all their choices arebetween losses. Money that is spent on one good is the loss of anothergood that could have been purchased instead. For the poor, costs arelosses.

We all know people for whom spending is painful, although they areobjectively quite well-off. There may also be cultural differences in theattitude toward money, and especially toward the spending of money onwhims Bon s Ahims Bon and minor luxuries, such as the purchase of adecorated mug. Such a difference may explain the large discrepancybetween the results of the “mugs study” in the United States and in the UK.Buying and selling prices diverge substantially in experiments conducted insamples of students of the United States, but the differences are muchsmaller among English students. Much remains to be learned about theendowment effect.

Speaking Of The Endowment Effect

“She didn’t care which of the two offices she would get, but a dayafter the announcement was made, she was no longer willing totrade. Endowment effect!”

“These negotiations are going nowhere because both sides findit difficult to make concessions, even when they can getsomething in return. Losses loom larger than gains.”

“When they raised their prices, demand dried up.”

“He just hates the idea of selling his house for less money than hepaid for it. Loss aversion is at work.”

(PDF) Thinking, fast and slow by daniel kahneman - PDFSLIDE.US (291)

“He is a miser, and treats any dollar he spends as a loss.”

(PDF) Thinking, fast and slow by daniel kahneman - PDFSLIDE.US (292)

Bad Events

The concept of loss aversion is certainly the most significant contribution ofpsychology to behavioral economics. This is odd, because the idea thatpeople evaluate many outcomes as gains and losses, and that lossesloom larger than gains, surprises no one. Amos and I often joked that wewere engaged in studying a subject about which our grandmothers knew agreat deal. In fact, however, we know more than our grandmothers did andcan now embed loss aversion in the context of a broader two-systemsmodel of the mind, and specifically a biological and psychological view inwhich negativity and escape dominate positivity and approach. We canalso trace the consequences of loss aversion in surprisingly diverseobservations: only out-of-pocket losses are compensated when goods arelost in transport; attempts at large-scale reforms very often fail; andprofessional golfers putt more accurately for par than for a birdie. Cleveras she was, my grandmother would have been surprised by the specificpredictions from a general idea she considered obvious.

Negativity Dominance

Figure 12

Your heartbeat accelerated when you looked at the left-hand figure. Itaccelerated even before you could label what is so eerie about thatpicture. After some time you may have recognized the eyes of a terrifiedperson. The eyes on the right, narrowed by the Crro raised cheeks of asmile, express happiness—and they are not nearly as exciting. The twopictures were presented to people lying in a brain scanner. Each picturewas shown for less than 2/100 of a second and immediately masked by“visual noise,” a random display of dark and bright squares. None of theobservers ever consciously knew that he had seen pictures of eyes, butone part of their brain evidently knew: the amygdala, which has a primaryrole as the “threat center” of the brain, although it is also activated in otheremotional states. Images of the brain showed an intense response of theamygdala to a threatening picture that the viewer did not recognize. The

(PDF) Thinking, fast and slow by daniel kahneman - PDFSLIDE.US (293)

information about the threat probably traveled via a superfast neuralchannel that feeds directly into a part of the brain that processes emotions,bypassing the visual cortex that supports the conscious experience of“seeing.” The same circuit also causes schematic angry faces (a potentialthreat) to be processed faster and more efficiently than schematic happyfaces. Some experimenters have reported that an angry face “pops out” ofa crowd of happy faces, but a single happy face does not stand out in anangry crowd. The brains of humans and other animals contain amechanism that is designed to give priority to bad news. By shaving a fewhundredths of a second from the time needed to detect a predator, thiscircuit improves the animal’s odds of living long enough to reproduce. Theautomatic operations of System 1 reflect this evolutionary history. Nocomparably rapid mechanism for recognizing good news has beendetected. Of course, we and our animal cousins are quickly alerted tosigns of opportunities to mate or to feed, and advertisers design billboardsaccordingly. Still, threats are privileged above opportunities, as they shouldbe.

The brain responds quickly even to purely symbolic threats. Emotionallyloaded words quickly attract attention, and bad words (war, crime) attractattention faster than do happy words (peace, love). There is no real threat,but the mere reminder of a bad event is treated in System 1 asthreatening. As we saw earlier with the word vomit, the symbolicrepresentation associatively evokes in attenuated form many of thereactions to the real thing, including physiological indices of emotion andeven fractional tendencies to avoid or approach, recoil or lean forward.The sensitivity to threats extends to the processing of statements ofopinions with which we strongly disagree. For example, depending on yourattitude to euthanasia, it would take your brain less than one-quarter of asecond to register the “threat” in a sentence that starts with “I thinkeuthanasia is an acceptable/unacceptable…”

The psychologist Paul Rozin, an expert on disgust, observed that asingle cockroach will completely wreck the appeal of a bowl of cherries,but a cherry will do nothing at all for a bowl of cockroaches. As he pointsout, the negative trumps the positive in many ways, and loss aversion isone of many manifestations of a broad negativity dominance. Otherscholars, in a paper titled “Bad Is Stronger Than Good,” summarized theevidence as follows: “Bad emotions, bad parents, and bad feedback havemore impact than good ones, and bad information is processed morethoroughly than good. The self is more motivated to avoid bad self-definitions than to pursue good ones. Bad impressions and badstereotypes are quicker to form and more resistant to disconfirmation than

(PDF) Thinking, fast and slow by daniel kahneman - PDFSLIDE.US (294)

good ones.” They cite John Gottman, the well-known expert in maritalrelations, who observed that the long-term success of a relationshipdepends far more on avoiding the negative than on seeking the positive.Gottman estimated that a stable relationship requires Brro Qres Brrthatgood interactions outnumber bad interactions by at least 5 to 1. Otherasymmetries in the social domain are even more striking. We all know thata friendship that may take years to develop can be ruined by a singleaction.

Some distinctions between good and bad are hardwired into ourbiology. Infants enter the world ready to respond to pain as bad and tosweet (up to a point) as good. In many situations, however, the boundarybetween good and bad is a reference point that changes over time anddepends on the immediate circumstances. Imagine that you are out in thecountry on a cold night, inadequately dressed for the torrential rain, yourclothes soaked. A stinging cold wind completes your misery. As youwander around, you find a large rock that provides some shelter from thefury of the elements. The biologist Michel Cabanac would call theexperience of that moment intensely pleasurable because it functions, aspleasure normally does, to indicate the direction of a biologicallysignificant improvement of circumstances. The pleasant relief will not lastvery long, of course, and you will soon be shivering behind the rock again,driven by your renewed suffering to seek better shelter.

Goals are Reference Points

Loss aversion refers to the relative strength of two motives: we are drivenmore strongly to avoid losses than to achieve gains. A reference point issometimes the status quo, but it can also be a goal in the future: notachieving a goal is a loss, exceeding the goal is a gain. As we mightexpect from negativity dominance, the two motives are not equallypowerful. The aversion to the failure of not reaching the goal is muchstronger than the desire to exceed it.

People often adopt short-term goals that they strive to achieve but notnecessarily to exceed. They are likely to reduce their efforts when theyhave reached an immediate goal, with results that sometimes violateeconomic logic. New York cabdrivers, for example, may have a targetincome for the month or the year, but the goal that controls their effort istypically a daily target of earnings. Of course, the daily goal is much easierto achieve (and exceed) on some days than on others. On rainy days, aNew York cab never remains free for long, and the driver quickly achieveshis target; not so in pleasant weather, when cabs often waste time cruising

(PDF) Thinking, fast and slow by daniel kahneman - PDFSLIDE.US (295)

the streets looking for fares. Economic logic implies that cabdrivers shouldwork many hours on rainy days and treat themselves to some leisure onmild days, when they can “buy” leisure at a lower price. The logic of lossaversion suggests the opposite: drivers who have a fixed daily target willwork many more hours when the pickings are slim and go home earlywhen rain-drenched customers are begging to be taken somewhere.

The economists Devin Pope and Maurice Schweitzer, at the Universityof Pennsylvania, reasoned that golf provides a perfect example of areference point: par. Every hole on the golf course has a number of strokesassociated with it; the par number provides the baseline for good—but notoutstanding—performance. For a professional golfer, a birdie (one strokeunder par) is a gain, and a bogey (one stroke over par) is a loss. Theeconomists compared two situations a player might face when near thehole:

putt to avoid a bogeyputt to achieve a birdie

Every stroke counts in golf, and in professional golf every stroke counts alot. According to prospect theory, however, some strokes count more thanothers. Failing to make par is a los Brro Q los Brrs, but missing a birdieputt is a foregone gain, not a loss. Pope and Schweitzer reasoned fromloss aversion that players would try a little harder when putting for par (toavoid a bogey) than when putting for a birdie. They analyzed more than 2.5million putts in exquisite detail to test that prediction.

They were right. Whether the putt was easy or hard, at every distancefrom the hole, the players were more successful when putting for par thanfor a birdie. The difference in their rate of success when going for par (toavoid a bogey) or for a birdie was 3.6%. This difference is not trivial. TigerWoods was one of the “participants” in their study. If in his best years TigerWoods had managed to putt as well for birdies as he did for par, hisaverage tournament score would have improved by one stroke and hisearnings by almost $1 million per season. These fierce competitorscertainly do not make a conscious decision to slack off on birdie putts, buttheir intense aversion to a bogey apparently contributes to extraconcentration on the task at hand.

The study of putts illustrates the power of a theoretical concept as an aidto thinking. Who would have thought it worthwhile to spend monthsanalyzing putts for par and birdie? The idea of loss aversion, which

(PDF) Thinking, fast and slow by daniel kahneman - PDFSLIDE.US (296)

surprises no one except perhaps some economists, generated a preciseand nonintuitive hypothesis and led researchers to a finding that surprisedeveryone—including professional golfers.

Defending the Status Quo

If you are set to look for it, the asymmetric intensity of the motives to avoidlosses and to achieve gains shows up almost everywhere. It is an ever-present feature of negotiations, especially of renegotiations of an existingcontract, the typical situation in labor negotiations and in internationaldiscussions of trade or arms limitations. The existing terms definereference points, and a proposed change in any aspect of the agreementis inevitably viewed as a concession that one side makes to the other.Loss aversion creates an asymmetry that makes agreements difficult toreach. The concessions you make to me are my gains, but they are yourlosses; they cause you much more pain than they give me pleasure.Inevitably, you will place a higher value on them than I do. The same is true,of course, of the very painful concessions you demand from me, which youdo not appear to value sufficiently! Negotiations over a shrinking pie areespecially difficult, because they require an allocation of losses. Peopletend to be much more easygoing when they bargain over an expandingpie.

Many of the messages that negotiators exchange in the course ofbargaining are attempts to communicate a reference point and provide ananchor to the other side. The messages are not always sincere.Negotiators often pretend intense attachment to some good (perhapsmissiles of a particular type in bargaining over arms reductions), althoughthey actually view that good as a bargaining chip and intend ultimately togive it away in an exchange. Because negotiators are influenced by anorm of reciprocity, a concession that is presented as painful calls for anequally painful (and perhaps equally inauthentic) concession from the otherside.

Animals, including people, fight harder to prevent losses than to achievegains. In the world of territorial animals, this principle explains the successof defenders. A biologist observed that “when a territory holder ischallenged by a rival, the owner almost always wins the contest—usuallywithin a matter of seconds.” In human affairs, the same simple rule explainsmuch of what happens when institutions attempt to reform themselves, in“reo Brro Q;reo Brrrganizations” and “restructuring” of companies, and inefforts to rationalize a bureaucracy, simplify the tax code, or reducemedical costs. As initially conceived, plans for reform almost always

(PDF) Thinking, fast and slow by daniel kahneman - PDFSLIDE.US (297)

produce many winners and some losers while achieving an overallimprovement. If the affected parties have any political influence, however,potential losers will be more active and determined than potential winners;the outcome will be biased in their favor and inevitably more expensiveand less effective than initially planned. Reforms commonly includegrandfather clauses that protect current stake-holders—for example, whenthe existing workforce is reduced by attrition rather than by dismissals, orwhen cuts in salaries and benefits apply only to future workers. Lossaversion is a powerful conservative force that favors minimal changes fromthe status quo in the lives of both institutions and individuals. Thisconservatism helps keep us stable in our neighborhood, our marriage, andour job; it is the gravitational force that holds our life together near thereference point.

Loss Aversion in the Law

During the year that we spent working together in Vancouver, RichardThaler, Jack Knetsch, and I were drawn into a study of fairness ineconomic transactions, partly because we were interested in the topic butalso because we had an opportunity as well as an obligation to make up anew questionnaire every week. The Canadian government’s Departmentof Fisheries and Oceans had a program for unemployed professionals inToronto, who were paid to administer telephone surveys. The large team ofinterviewers worked every night and new questions were constantlyneeded to keep the operation going. Through Jack Knetsch, we agreed togenerate a questionnaire every week, in four color-labeled versions. Wecould ask about anything; the only constraint was that the questionnaireshould include at least one mention of fish, to make it pertinent to themission of the department. This went on for many months, and we treatedourselves to an orgy of data collection.

We studied public perceptions of what constitutes unfair behavior on thepart of merchants, employers, and landlords. Our overarching questionwas whether the opprobrium attached to unfairness imposes constraintson profit seeking. We found that it does. We also found that the moral rulesby which the public evaluates what firms may or may not do draw a crucialdistinction between losses and gains. The basic principle is that theexisting wage, price, or rent sets a reference point, which has the nature ofan entitlement that must not be infringed. It is considered unfair for the firmto impose losses on its customers or workers relative to the referencetransaction, unless it must do so to protect its own entitlement. Considerthis example:

(PDF) Thinking, fast and slow by daniel kahneman - PDFSLIDE.US (298)

A hardware store has been selling snow shovels for $15. Themorning after a large snowstorm, the store raises the price to$20.Please rate this action as:Completely Fair Acceptable Unfair Very Unfair

The hardware store behaves appropriately according to the standardeconomic model: it responds to increased demand by raising its price.The participants in the survey did not agree: 82% rated the action Unfair orVery Unfair. They evidently viewed the pre-blizzard price as a referencepoint and the raised price as a loss that the store imposes on itscustomers, not because it must but simply because it can. A basic rule offairness, we found, i Brro Qd, i Brrs that the exploitation of market power toimpose losses on others is unacceptable. The following example illustratesthis rule in another context (the dollar values should be adjusted for about100% inflation since these data were collected in 1984):

A small photocopying shop has one employee who has workedthere for six months and earns $9 per hour. Business continues tobe satisfactory, but a factory in the area has closed andunemployment has increased. Other small shops have now hiredreliable workers at $7 an hour to perform jobs similar to thosedone by the photocopy shop employee. The owner of the shopreduces the employee’s wage to $7.

The respondents did not approve: 83% considered the behavior Unfair orVery Unfair. However, a slight variation on the question clarifies the natureof the employer’s obligation. The background scenario of a profitable storein an area of high unemployment is the same, but now

the current employee leaves, and the owner decides to pay areplacement $7 an hour.

A large majority (73%) considered this action Acceptable. It appears thatthe employer does not have a moral obligation to pay $9 an hour. Theentitlement is personal: the current worker has a right to retain his wageeven if market conditions would allow the employer to impose a wage cut.The replacement worker has no entitlement to the previous worker’sreference wage, and the employer is therefore allowed to reduce paywithout the risk of being branded unfair.

The firm has its own entitlement, which is to retain its current profit. If itfaces a threat of a loss, it is allowed to transfer the loss to others. A

(PDF) Thinking, fast and slow by daniel kahneman - PDFSLIDE.US (299)

substantial majority of respondents believed that it is not unfair for a firm toreduce its workers’ wages when its profitability is falling. We described therules as defining dual entitlements to the firm and to individuals with whomit interacts. When threatened, it is not unfair for the firm to be selfish. It isnot even expected to take on part of the losses; it can pass them on.

Different rules governed what the firm could do to improve its profits orto avoid reduced profits. When a firm faced lower production costs, therules of fairness did not require it to share the bonanza with either itscustomers or its workers. Of course, our respondents liked a firm betterand described it as more fair if it was generous when its profits increased,but they did not brand as unfair a firm that did not share. They showedindignation only when a firm exploited its power to break informal contractswith workers or customers, and to impose a loss on others in order toincrease its profit. The important task for students of economic fairness isnot to identify ideal behavior but to find the line that separates acceptableconduct from actions that invite opprobrium and punishment.

We were not optimistic when we submitted our report of this research tothe American Economic Review. Our article challenged what was thenaccepted wisdom among many economists that economic behavior isruled by self-interest and that concerns for fairness are generally irrelevant.We also relied on the evidence of survey responses, for which economistsgenerally have little respect. However, the editor of the journal sent ourarticle for evaluation to two economists who were not bound by thoseconventions (we later learned their identity; they were the most friendly theeditor could have found). The editor made the correct call. The article isoften cited, and its conclusions Brro Qions Brr have stood the test of time.More recent research has supported the observations of reference-dependent fairness and has also shown that fairness concerns areeconomically significant, a fact we had suspected but did not prove.Employers who violate rules of fairness are punished by reducedproductivity, and merchants who follow unfair pricing policies can expect tolose sales. People who learned from a new catalog that the merchant wasnow charging less for a product that they had recently bought at a higherprice reduced their future purchases from that supplier by 15%, an averageloss of $90 per customer. The customers evidently perceived the lowerprice as the reference point and thought of themselves as having sustaineda loss by paying more than appropriate. Moreover, the customers whoreacted the most strongly were those who bought more items and at higherprices. The losses far exceeded the gains from the increased purchasesproduced by the lower prices in the new catalog.

Unfairly imposing losses on people can be risky if the victims are in aposition to retaliate. Furthermore, experiments have shown that strangers

(PDF) Thinking, fast and slow by daniel kahneman - PDFSLIDE.US (300)

position to retaliate. Furthermore, experiments have shown that strangerswho observe unfair behavior often join in the punishment.Neuroeconomists (scientists who combine economics with brain research)have used MRI machines to examine the brains of people who areengaged in punishing one stranger for behaving unfairly to anotherstranger. Remarkably, altruistic punishment is accompanied by increasedactivity in the “pleasure centers” of the brain. It appears that maintaining thesocial order and the rules of fairness in this fashion is its own reward.Altruistic punishment could well be the glue that holds societies together.However, our brains are not designed to reward generosity as reliably asthey punish meanness. Here again, we find a marked asymmetry betweenlosses and gains.

The influence of loss aversion and entitlements extends far beyond therealm of financial transactions. Jurists were quick to recognize their impacton the law and in the administration of justice. In one study, David Cohenand Jack Knetsch found many examples of a sharp distinction betweenactual losses and foregone gains in legal decisions. For example, amerchant whose goods were lost in transit may be compensated for costshe actually incurred, but is unlikely to be compensated for lost profits. Thefamiliar rule that possession is nine-tenths of the law confirms the moralstatus of the reference point. In a more recent discussion, Eyal Zamirmakes the provocative point that the distinction drawn in the law betweenrestoring losses and compensating for foregone gains may be justified bytheir asymmetrical effects on individual well-being. If people who losesuffer more than people who merely fail to gain, they may also deservemore protection from the law.

Speaking of Losses

“This reform will not pass. Those who stand to lose will fightharder than those who stand to gain.”

“Each of them thinks the other’s concessions are less painful.They are both wrong, of course. It’s just the asymmetry of losses.”

“They would find it easier to renegotiate the agreement if theyrealized the pie was actually expanding. They’re not allocatinglosses; they are allocating gains.”

(PDF) Thinking, fast and slow by daniel kahneman - PDFSLIDE.US (301)

“Rental prices around here have gone up r Brro Qup r Brrecently,but our tenants don’t think it’s fair that we should raise their rent,too. They feel entitled to their current terms.”

“My clients don’t resent the price hike because they know mycosts have gone up, too. They accept my right to stay profitable.”

(PDF) Thinking, fast and slow by daniel kahneman - PDFSLIDE.US (302)

The Fourfold Pattern

Whenever you form a global evaluation of a complex object—a car youmay buy, your son-in-law, or an uncertain situation—you assign weights toits characteristics. This is simply a cumbersome way of saying that somecharacteristics influence your assessment more than others do. Theweighting occurs whether or not you are aware of it; it is an operation ofSystem 1. Your overall evaluation of a car may put more or less weight ongas economy, comfort, or appearance. Your judgment of your son-in-lawmay depend more or less on how rich or handsome or reliable he is.Similarly, your assessment of an uncertain prospect assigns weights to thepossible outcomes. The weights are certainly correlated with theprobabilities of these outcomes: a 50% chance to win a million is muchmore attractive than a 1% chance to win the same amount. Theassignment of weights is sometimes conscious and deliberate. Most often,however, you are just an observer to a global evaluation that your System 1delivers.

Changing Chances

One reason for the popularity of the gambling metaphor in the study ofdecision making is that it provides a natural rule for the assignment ofweights to the outcomes of a prospect: the more probable an outcome, themore weight it should have. The expected value of a gamble is the averageof its outcomes, each weighted by its probability. For example, theexpected value of “20% chance to win $1,000 and 75% chance to win$100” is $275. In the pre-Bernoulli days, gambles were assessed by theirexpected value. Bernoulli retained this method for assigning weights to theoutcomes, which is known as the expectation principle, but applied it to thepsychological value of the outcomes. The utility of a gamble, in his theory,is the average of the utilities of its outcomes, each weighted by itsprobability.

The expectation principle does not correctly describe how you thinkabout the probabilities related to risky prospects. In the four examplesbelow, your chances of receiving $1 million improve by 5%. Is the newsequally good in each case?

A. From 0 to 5%B. From 5% to 10%C. From 60% to 65%D. From 95% to 100%

(PDF) Thinking, fast and slow by daniel kahneman - PDFSLIDE.US (303)

The expectation principle asserts that your utility increases in each case byexactly 5% of the utility of receiving $1 million. Does this predictiondescribe your experiences? Of course not.

Everyone agrees that 0 5% and 95% 100% are more impressivethan either 5% 10% or 60% 65%. Increasing the chances from 0 to5% transforms the situation, creating a possibility that did not exist earlier,a hope of winning the prize. It is a qualitative change, where 5 10% isonly a quantitative improvement. The change from 5% to 10% doubles theprobability of winning, but there is general agreement that thepsychological value of the prospect does not double. The large impact of 0

5% illustrates the possibility effect, which causes highly unlikelyoutcomes to be weighted disproportionately more than they “deserve.”People who buy lottery tickets in vast amounts show themselves willing topay much more than expected value for very small chances to win a largeprize.

The improvement from 95% to 100% is another qualitative change thathas a large impact, the certainty effect. Outcomes that are almost certainare given less weight than their probability justifies. To appreciate thecertainty effect, imagine that you inherited $1 million, but your greedystepsister has contested the will in court. The decision is expectedtomorrow. Your lawyer assures you that you have a strong case and thatyou have a 95% chance to win, but he takes pains to remind you thatjudicial decisions are never perfectly predictable. Now you areapproached by a risk-adjustment company, which offers to buy your casefor $910,000 outright—take it or leave it. The offer is lower (by $40,000!)than the expected value of waiting for the judgment (which is $950,000),but are you quite sure you would want to reject it? If such an event actuallyhappens in your life, you should know that a large industry of “structuredsettlements” exists to provide certainty at a heft y price, by takingadvantage of the certainty effect.

Possibility and certainty have similarly powerful effects in the domain oflosses. When a loved one is wheeled into surgery, a 5% risk that anamputation will be necessary is very bad—much more than half as bad asa 10% risk. Because of the possibility effect, we tend to overweight smallrisks and are willing to pay far more than expected value to eliminate themaltogether. The psychological difference between a 95% risk of disasterand the certainty of disaster appears to be even greater; the sliver of hopethat everything could still be okay looms very large. Overweighting of smallprobabilities increases the attractiveness of both gambles and insurancepolicies.

(PDF) Thinking, fast and slow by daniel kahneman - PDFSLIDE.US (304)

The conclusion is straightforward: the decision weights that peopleassign to outcomes are not identical to the probabilities of theseoutcomes, contrary to the expectation principle. Improbable outcomes areoverweighted—this is the possibility effect. Outcomes that are almostcertain are underweighted relative to actual certainty. The expectationprinciple, by which values are weighted by their probability, is poorpsychology.

The plot thickens, however, because there is a powerful argument that adecision maker who wishes to be rational must conform to the expectationprinciple. This was the main point of the axiomatic version of utility theorythat von Neumann and Morgenstern introduced in 1944. They proved thatany weighting of uncertain outcomes that is not strictly proportional toprobability leads to inconsistencies and other disasters. Their derivation ofthe expectation principle from axioms of rational choice was immediatelyrecognized as a monumental achievement, which placed expected utilitytheory at the core of the rational agent model in economics and othersocial sciences. Thirty years later, when Amos introduced me to their work,he presented it as an object of awe. He also introduced me Bima a meBimto a famous challenge to that theory.

Allais’s Paradox

In 1952, a few years after the publication of von Neumann andMorgenstern’s theory, a meeting was convened in Paris to discuss theeconomics of risk. Many of the most renowned economists of the timewere in attendance. The American guests included the future Nobellaureates Paul Samuelson, Kenneth Arrow, and Milton Friedman, as wellas the leading statistician Jimmie Savage.

One of the organizers of the Paris meeting was Maurice Allais, whowould also receive a Nobel Prize some years later. Allais had somethingup his sleeve, a couple of questions on choice that he presented to hisdistinguished audience. In the terms of this chapter, Allais intended toshow that his guests were susceptible to a certainty effect and thereforeviolated expected utility theory and the axioms of rational choice on whichthat theory rests. The following set of choices is a simplified version of thepuzzle that Allais constructed. In problems A and B, which would youchoose?

A. 61% chance to win $520,000 OR 63% chance to win $500,000

B. 98% chance to win $520,000 OR 100% chance to win $500,000

(PDF) Thinking, fast and slow by daniel kahneman - PDFSLIDE.US (305)

If you are like most other people, you preferred the left-hand option inproblem A and you preferred the right-hand option in problem B. If thesewere your preferences, you have just committed a logical sin and violatedthe rules of rational choice. The illustrious economists assembled in Pariscommitted similar sins in a more involved version of the “Allais paradox.”

To see why these choices are problematic, imagine that the outcomewill be determined by a blind draw from an urn that contains 100 marbles—you win if you draw a red marble, you lose if you draw white. In problem A,almost everybody prefers the left-hand urn, although it has fewer winningred marbles, because the difference in the size of the prize is moreimpressive than the difference in the chances of winning. In problem B, alarge majority chooses the urn that guarantees a gain of $500,000.Furthermore, people are comfortable with both choices—until they are ledthrough the logic of the problem.

Compare the two problems, and you will see that the two urns ofproblem B are more favorable versions of the urns of problem A, with 37white marbles replaced by red winning marbles in each urn. Theimprovement on the left is clearly superior to the improvement on the right,since each red marble gives you a chance to win $520,000 on the left andonly $500,000 on the right. So you started in the first problem with apreference for the left-hand urn, which was then improved more than theright-hand urn—but now you like the one on the right! This pattern ofchoices does not make logical sense, but a psychological explanation isreadily available: the certainty effect is at work. The 2% difference betweena 100% and a 98% chance to win in problem B is vastly more impressivethan the same difference between 63% and 61% in problem A.

As Allais had anticipated, the sophisticated participants at the meetingdid not notice that their preferences violated utility theory until he drew theirattention to that fact as the meeting was about to end. Allais had intendedthis announcement to be a bombshell: the leading decision theorists in theworld had preferences that were inconsistent with their own view ofrationality! He apparently believed that his audience would be persuadedto give up the approach that Bima ahat Bimhe rather contemptuouslylabeled “the American school” and adopt an alternative logic of choice thathe had developed. He was to be sorely disappointed.

Economists who were not aficionados of decision theory mostly ignoredthe Allais problem. As often happens when a theory that has been widelyadopted and found useful is challenged, they noted the problem as ananomaly and continued using expected utility theory as if nothing hadhappened. In contrast, decision theorists—a mixed collection of

(PDF) Thinking, fast and slow by daniel kahneman - PDFSLIDE.US (306)

statisticians, economists, philosophers, and psychologists—took Allais’schallenge very seriously. When Amos and I began our work, one of ourinitial goals was to develop a satisfactory psychological account of Allais’sparadox.

Most decision theorists, notably including Allais, maintained their beliefin human rationality and tried to bend the rules of rational choice to makethe Allais pattern permissible. Over the years there have been multipleattempts to find a plausible justification for the certainty effect, none veryconvincing. Amos had little patience for these efforts; he called thetheorists who tried to rationalize violations of utility theory “lawyers for themisguided.” We went in another direction. We retained utility theory as alogic of rational choice but abandoned the idea that people are perfectlyrational choosers. We took on the task of developing a psychologicaltheory that would describe the choices people make, regardless ofwhether they are rational. In prospect theory, decision weights would not beidentical to probabilities.

Decision Weights

Many years after we published prospect theory, Amos and I carried out astudy in which we measured the decision weights that explained people’spreferences for gambles with modest monetary stakes. The estimates forgains are shown in table 4.

Table 4

You can see that the decision weights are identical to the correspondingprobabilities at the extremes: both equal to 0 when the outcome isimpossible, and both equal to 100 when the outcome is a sure thing.However, decision weights depart sharply from probabilities near thesepoints. At the low end, we find the possibility effect: unlikely events areconsiderably overweighted. For example, the decision weight thatcorresponds to a 2% chance is 8.1. If people conformed to the axioms ofrational choice, the decision weight would be 2—so the rare event isoverweighted by a factor of 4. The certainty effect at the other end of theprobability scale is even more striking. A 2% risk of not winning the prizereduces the utility of the gamble by 13%, from 100 to 87.1.

To appreciate the asymmetry between the possibility effect and the

(PDF) Thinking, fast and slow by daniel kahneman - PDFSLIDE.US (307)

certainty effect, imagine first that you have a 1% chance to win $1 million.You will know the outcome tomorrow. Now, imagine that you are almostcertain to win $1 million, but there is a 1% chance that you will not. Again,you will learn the outcome tomorrow. The anxiety of the second situationappears to be more salient than the hope in the first. The certainty effect isalso more striking than the possibility effect if the outcome is a surgicaldisaster rather than a financial gain. Compare the intensity with which youfocus on the faint sliver of hope in an operation that is almost certain to befatal, compared to the fear of a 1% risk.< Bima av> < Bimp height="0%" width="5%">The combination of thecertainty effect and possibility effects at the two ends of the probabilityscale is inevitably accompanied by inadequate sensitivity to intermediateprobabilities. You can see that the range of probabilities between 5% and95% is associated with a much smaller range of decision weights (from13.2 to 79.3), about two-thirds as much as rationally expected.Neuroscientists have confirmed these observations, finding regions of thebrain that respond to changes in the probability of winning a prize. Thebrain’s response to variations of probabilities is strikingly similar to thedecision weights estimated from choices.

Probabilities that are extremely low or high (below 1% or above 99%)are a special case. It is difficult to assign a unique decision weight to veryrare events, because they are sometimes ignored altogether, effectivelyassigned a decision weight of zero. On the other hand, when you do notignore the very rare events, you will certainly overweight them. Most of usspend very little time worrying about nuclear meltdowns or fantasizingabout large inheritances from unknown relatives. However, when anunlikely event becomes the focus of attention, we will assign it much moreweight than its probability deserves. Furthermore, people are almostcompletely insensitive to variations of risk among small probabilities. Acancer risk of 0.001% is not easily distinguished from a risk of 0.00001%,although the former would translate to 3,000 cancers for the population ofthe United States, and the latter to 30.

When you pay attention to a threat, you worry—and the decision weightsreflect how much you worry. Because of the possibility effect, the worry isnot proportional to the probability of the threat. Reducing or mitigating therisk is not adequate; to eliminate the worry the probability must be broughtdown to zero.

The question below is adapted from a study of the rationality ofconsumer valuations of health risks, which was published by a team ofeconomists in the 1980s. The survey was addressed to parents of small

(PDF) Thinking, fast and slow by daniel kahneman - PDFSLIDE.US (308)

children.

Suppose that you currently use an insect spray that costs you $10per bottle and it results in 15 inhalation poisonings and 15 childpoisonings for every 10,000 bottles of insect spray that are used.

You learn of a more expensive insecticide that reduces each ofthe risks to 5 for every 10,000 bottles. How much would you bewilling to pay for it?

The parents were willing to pay an additional $2.38, on average, to reducethe risks by two-thirds from 15 per 10,000 bottles to 5. They were willing topay $8.09, more than three times as much, to eliminate it completely. Otherquestions showed that the parents treated the two risks (inhalation andchild poisoning) as separate worries and were willing to pay a certaintypremium for the complete elimination of either one. This premium iscompatible with the psychology of worry but not with the rational model.

The Fourfold Pattern

When Amos and I began our work on prospect theory, we quickly reachedtwo conclusions: people attach values to gains and losses rather than towealth, and the decision weights that they assign to outcomes are differentfrom probabilities. Neither idea was completely new, but in combinationthey explained a distinctive pattern of preferences that we ca Bima ae caBimlled the fourfold pattern. The name has stuck. The scenarios areillustrated below.

(PDF) Thinking, fast and slow by daniel kahneman - PDFSLIDE.US (309)

Figure 13

The top row in each cell shows an illustrative prospect.The second row characterizes the focal emotion that the prospectevokes.The third row indicates how most people behave when offered achoice between a gamble and a sure gain (or loss) that correspondsto its expected value (for example, between “95% chance to win$10,000” and “$9,500 with certainty”). Choices are said to be riskaverse if the sure thing is preferred, risk seeking if the gamble ispreferred.The fourth row describes the expected attitudes of a defendant and aplaintiff as they discuss a settlement of a civil suit.

T he fourfold pattern of preferences is considered one of the coreachievements of prospect theory. Three of the four cells are familiar; thefourth (top right) was new and unexpected.

The top left is the one that Bernoulli discussed: people are averse torisk when they consider prospects with a substantial chance toachieve a large gain. They are willing to accept less than theexpected value of a gamble to lock in a sure gain.The possibility effect in the bottom left cell explains why lotteries arepopular. When the top prize is very large, ticket buyers appearindifferent to the fact that their chance of winning is minuscule. Alottery ticket is the ultimate example of the possibility effect. Withouta ticket you cannot win, with a ticket you have a chance, and whetherthe chance is tiny or merely small matters little. Of course, whatpeople acquire with a ticket is more than a chance to win; it is theright to dream pleasantly of winning.The bottom right cell is where insurance is bought. People are willingto pay much more for insurance than expected value—which is howinsurance companies cover their costs and make their profits. Hereagain, people buy more than protection against an unlikely disaster;they eliminate a worry and purchase peace of mind.

(PDF) Thinking, fast and slow by daniel kahneman - PDFSLIDE.US (310)

The results for the top right cell initially surprised us. We were accustomedto think in terms of risk aversion except for the bottom left cell, wherelotteries are preferred. When we looked at our choices for bad options, wequickly realized that we were just as risk seeking in the domain of lossesas we were risk averse in the domain of gains. We were not the first toobserve risk seeking with negative prospects—at least two authors hadreported that fact, but they had not made much of it. However, we werefortunate to have a framework that made the finding of risk seeking easy tointerpret, and that was a milestone in our thinking. Indeed, we identifiedtwo reasons for this effect.

First, there is diminishing sensitivity. The sure loss is very aversivebecause the reaction to a loss of $900 is more than 90% as intense as thereaction to a loss of $1,000. The second factor may be even morepowerful: the decision weight that corresponds to a probability of 90% isonly about 71, much lower than the probability. The result is that when youconsider a choice between a sure loss and a gamble with a highprobability o Bima aty o Bimf a larger loss, diminishing sensitivity makesthe sure loss more aversive, and the certainty effect reduces theaversiveness of the gamble. The same two factors enhance theattractiveness of the sure thing and reduce the attractiveness of thegamble when the outcomes are positive.

The shape of the value function and the decision weights both contributeto the pattern observed in the top row of table 13. In the bottom row,however, the two factors operate in opposite directions: diminishingsensitivity continues to favor risk aversion for gains and risk seeking forlosses, but the overweighting of low probabilities overcomes this effectand produces the observed pattern of gambling for gains and caution forlosses.

Many unfortunate human situations unfold in the top right cell. This iswhere people who face very bad options take desperate gambles,accepting a high probability of making things worse in exchange for asmall hope of avoiding a large loss. Risk taking of this kind often turnsmanageable failures into disasters. The thought of accepting the large sureloss is too painful, and the hope of complete relief too enticing, to make thesensible decision that it is time to cut one’s losses. This is wherebusinesses that are losing ground to a superior technology waste theirremaining assets in futile attempts to catch up. Because defeat is sodifficult to accept, the losing side in wars often fights long past the point atwhich the victory of the other side is certain, and only a matter of time.

(PDF) Thinking, fast and slow by daniel kahneman - PDFSLIDE.US (311)

Gambling in the Shadow of the Law

The legal scholar Chris Guthrie has offered a compelling application of thefourfold pattern to two situations in which the plaintiff and the defendant in acivil suit consider a possible settlement. The situations differ in the strengthof the plaintiff’s case.

As in a scenario we saw earlier, you are the plaintiff in a civil suit inwhich you have made a claim for a large sum in damages. The trial isgoing very well and your lawyer cites expert opinion that you have a 95%chance to win outright, but adds the caution, “You never really know theoutcome until the jury comes in.” Your lawyer urges you to accept asettlement in which you might get only 90% of your claim. You are in the topleft cell of the fourfold pattern, and the question on your mind is, “Am Iwilling to take even a small chance of getting nothing at all? Even 90% ofthe claim is a great deal of money, and I can walk away with it now.” Twoemotions are evoked, both driving in the same direction: the attraction of asure (and substantial) gain and the fear of intense disappointment andregret if you reject a settlement and lose in court. You can feel the pressurethat typically leads to cautious behavior in this situation. The plaintiff with astrong case is likely to be risk averse.

Now step into the shoes of the defendant in the same case. Althoughyou have not completely given up hope of a decision in your favor, yourealize that the trial is going poorly. The plaintiff’s lawyers have proposed asettlement in which you would have to pay 90% of their original claim, andit is clear they will not accept less. Will you settle, or will you pursue thecase? Because you face a high probability of a loss, your situation belongsin the top right cell. The temptation to fight on is strong: the settlement thatthe plaintiff has offered is almost as painful as the worst outcome you face,and there is still hope of prevailing in court. Here again, two emotions areinvolved: the sure loss is repugnant and the possibility of winning in court ishighly attractive. A defendant with a weak case is likely to be risk seeking,Bima aing, Bim prepared to gamble rather than accept a very unfavorablesettlement. In the face-off between a risk-averse plaintiff and a risk-seekingdefendant, the defendant holds the stronger hand. The superior bargainingposition of the defendant should be reflected in negotiated settlements,with the plaintiff settling for less than the statistically expected outcome ofthe trial. This prediction from the fourfold pattern was confirmed byexperiments conducted with law students and practicing judges, and alsoby analyses of actual negotiations in the shadow of civil trials.

Now consider “frivolous litigation,” when a plaintiff with a flimsy case filesa large claim that is most likely to fail in court. Both sides are aware of the

(PDF) Thinking, fast and slow by daniel kahneman - PDFSLIDE.US (312)

probabilities, and both know that in a negotiated settlement the plaintiff willget only a small fraction of the amount of the claim. The negotiation isconducted in the bottom row of the fourfold pattern. The plaintiff is in theleft-hand cell, with a small chance to win a very large amount; the frivolousclaim is a lottery ticket for a large prize. Overweighting the small chance ofsuccess is natural in this situation, leading the plaintiff to be bold andaggressive in the negotiation. For the defendant, the suit is a nuisance witha small risk of a very bad outcome. Overweighting the small chance of alarge loss favors risk aversion, and settling for a modest amount isequivalent to purchasing insurance against the unlikely event of a badverdict. The shoe is now on the other foot: the plaintiff is willing to gambleand the defendant wants to be safe. Plaintiffs with frivolous claims arelikely to obtain a more generous settlement than the statistics of thesituation justify.

The decisions described by the fourfold pattern are not obviouslyunreasonable. You can empathize in each case with the feelings of theplaintiff and the defendant that lead them to adopt a combative or anaccommodating posture. In the long run, however, deviations fromexpected value are likely to be costly. Consider a large organization, theCity of New York, and suppose it faces 200 “frivolous” suits each year,each with a 5% chance to cost the city $1 million. Suppose further that ineach case the city could settle the lawsuit for a payment of $100,000. Thecity considers two alternative policies that it will apply to all such cases:settle or go to trial. (For simplicity, I ignore legal costs.)

If the city litigates all 200 cases, it will lose 10, for a total loss of $10million.If the city settles every case for $100,000, its total loss will be $20million.

When you take the long view of many similar decisions, you can see thatpaying a premium to avoid a small risk of a large loss is costly. A similaranalysis applies to each of the cells of the fourfold pattern: systematicdeviations from expected value are costly in the long run—and this ruleapplies to both risk aversion and risk seeking. Consistent overweighting ofimprobable outcomes—a feature of intuitive decision making—eventuallyleads to inferior outcomes.

Speaking Of The Fourfold Pattern

(PDF) Thinking, fast and slow by daniel kahneman - PDFSLIDE.US (313)

“He is tempted to settle this frivolous claim to avoid a freak loss,however unlikely. That’s overweighting of small probabilities.Since he is likely to face many similar problems, he would bebetter off not yielding.”

“We never let our vacations hang Bima aang Bimon a last-minutedeal. We’re willing to pay a lot for certainty.”

“They will not cut their losses so long as there is a chance ofbreaking even. This is risk-seeking in the losses.”

“They know the risk of a gas explosion is minuscule, but they wantit mitigated. It’s a possibility effect, and they want peace of mind.”

(PDF) Thinking, fast and slow by daniel kahneman - PDFSLIDE.US (314)

Rare Events

I visited Israel several times during a period in which suicide bombings inbuses were relatively common—though of course quite rare in absoluteterms. There were altogether 23 bombings between December 2001 andSeptember 2004, which had caused a total of 236 fatalities. The number ofdaily bus riders in Israel was approximately 1.3 million at that time. For anytraveler, the risks were tiny, but that was not how the public felt about it.People avoided buses as much as they could, and many travelers spenttheir time on the bus anxiously scanning their neighbors for packages orbulky clothes that might hide a bomb.

I did not have much occasion to travel on buses, as I was driving arented car, but I was chagrined to discover that my behavior was alsoaffected. I found that I did not like to stop next to a bus at a red light, and Idrove away more quickly than usual when the light changed. I wasashamed of myself, because of course I knew better. I knew that the riskwas truly negligible, and that any effect at all on my actions would assign aninordinately high “decision weight” to a minuscule probability. In fact, I wasmore likely to be injured in a driving accident than by stopping near a bus.But my avoidance of buses was not motivated by a rational concern forsurvival. What drove me was the experience of the moment: being next to abus made me think of bombs, and these thoughts were unpleasant. I wasavoiding buses because I wanted to think of something else.

My experience illustrates how terrorism works and why it is so effective:it induces an availability cascade. An extremely vivid image of death anddamage, constantly reinforced by media attention and frequentconversations, becomes highly accessible, especially if it is associatedwith a specific situation such as the sight of a bus. The emotional arousalis associative, automatic, and uncontrolled, and it produces an impulse forprotective action. System 2 may “know” that the probability is low, but thisknowledge does not eliminate the self-generated discomfort and the wishto avoid it. System 1 cannot be turned off. The emotion is not onlydisproportionate to the probability, it is also insensitive to the exact level ofprobability. Suppose that two cities have been warned about the presenceof suicide bombers. Residents of one city are told that two bombers areready to strike. Residents of another city are told of a single bomber. Theirrisk is lower by half, but do they feel much safer?

Many stores in New York City sell lottery tickets, and business is good. Thepsychology of high-prize lotteries is similar to the psychology of terrorism.

(PDF) Thinking, fast and slow by daniel kahneman - PDFSLIDE.US (315)

The thrilling possibility of winning the big prize is shared by the communityand re Cmuninforced by conversations at work and at home. Buying aticket is immediately rewarded by pleasant fantasies, just as avoiding abus was immediately rewarded by relief from fear. In both cases, the actualprobability is inconsequential; only possibility matters. The originalformulation of prospect theory included the argument that “highly unlikelyevents are either ignored or overweighted,” but it did not specify theconditions under which one or the other will occur, nor did it propose apsychological interpretation of it. My current view of decision weights hasbeen strongly influenced by recent research on the role of emotions andvividness in decision making. Overweighting of unlikely outcomes is rootedin System 1 features that are familiar by now. Emotion and vividnessinfluence fluency, availability, and judgments of probability—and thusaccount for our excessive response to the few rare events that we do notignore.

Overestimation and Overweighting

What is your judgment of the probability that the next president ofthe United States will be a third-party candidate?

How much will you pay for a bet in which you receive $1,000 if thenext president of the United States is a third-party candidate, andno money otherwise?

The two questions are different but obviously related. The first asks you toassess the probability of an unlikely event. The second invites you to put adecision weight on the same event, by placing a bet on it.

How do people make the judgments and how do they assign decisionweights? We start from two simple answers, then qualify them. Here arethe oversimplified answers:

People overestimate the probabilities of unlikely events.People overweight unlikely events in their decisions.

Although overestimation and overweighting are distinct phenomena, thesame psychological mechanisms are involved in both: focused attention,

(PDF) Thinking, fast and slow by daniel kahneman - PDFSLIDE.US (316)

confirmation bias, and cognitive ease.Specific descriptions trigger the associative machinery of System 1.

When you thought about the unlikely victory of a third-party candidate, yourassociative system worked in its usual confirmatory mode, selectivelyretrieving evidence, instances, and images that would make the statementtrue. The process was biased, but it was not an exercise in fantasy. Youlooked for a plausible scenario that conforms to the constraints of reality;you did not simply imagine the Fairy of the West installing a third-partypresident. Your judgment of probability was ultimately determined by thecognitive ease, or fluency, with which a plausible scenario came to mind.

You do not always focus on the event you are asked to estimate. If thetarget event is very likely, you focus on its alternative. Consider thisexample:

What is the probability that a baby born in your local hospital willbe released within three days?

You were asked to estimate the probability of the baby going home, butyou almost certainly focused on the events that might cause a baby not tobe released within the normal period. Our mind has a useful capability toBmun q to Bmufocus spontaneously on whatever is odd, different, orunusual. You quickly realized that it is normal for babies in the UnitedStates (not all countries have the same standards) to be released withintwo or three days of birth, so your attention turned to the abnormalalternative. The unlikely event became focal. The availability heuristic islikely to be evoked: your judgment was probably determined by the numberof scenarios of medical problems you produced and by the ease withwhich they came to mind. Because you were in confirmatory mode, there isa good chance that your estimate of the frequency of problems was toohigh.

The probability of a rare event is most likely to be overestimated whenthe alternative is not fully specified. My favorite example comes from astudy that the psychologist Craig Fox conducted while he was Amos’sstudent. Fox recruited fans of professional basketball and elicited severaljudgments and decisions concerning the winner of the NBA playoffs. Inparticular, he asked them to estimate the probability that each of the eightparticipating teams would win the playoff; the victory of each team in turnwas the focal event.

You can surely guess what happened, but the magnitude of the effectthat Fox observed may surprise you. Imagine a fan who has been asked toestimate the chances that the Chicago Bulls will win the tournament. Thefocal event is well defined, but its alternative—one of the other seven

(PDF) Thinking, fast and slow by daniel kahneman - PDFSLIDE.US (317)

teams winning—is diffuse and less evocative. The fan’s memory andimagination, operating in confirmatory mode, are trying to construct avictory for the Bulls. When the same person is next asked to assess thechances of the Lakers, the same selective activation will work in favor ofthat team. The eight best professional basketball teams in the UnitedStates are all very good, and it is possible to imagine even a relativelyweak team among them emerging as champion. The result: the probabilityjudgments generated successively for the eight teams added up to 240%!This pattern is absurd, of course, because the sum of the chances of theeight events must add up to 100%. The absurdity disappeared when thesame judges were asked whether the winner would be from the Eastern orthe Western conference. The focal event and its alternative were equallyspecific in that question and the judgments of their probabilities added upto 100%.

To assess decision weights, Fox also invited the basketball fans to beton the tournament result. They assigned a cash equivalent to each bet (acash amount that was just as attractive as playing the bet). Winning the betwould earn a payoff of $160. The sum of the cash equivalents for the eightindividual teams was $287. An average participant who took all eight betswould be guaranteed a loss of $127! The participants surely knew thatthere were eight teams in the tournament and that the average payoff forbetting on all of them could not exceed $160, but they overweightednonetheless. The fans not only overestimated the probability of the eventsthey focused on—they were also much too willing to bet on them.

These findings shed new light on the planning fallacy and othermanifestations of optimism. The successful execution of a plan is specificand easy to imagine when one tries to forecast the outcome of a project. Incontrast, the alternative of failure is diffuse, because there are innumerableways for things to go wrong. Entrepreneurs and the investors who evaluatetheir prospects are prone both to overestimate their chances and tooverweight their estimates.

Vivid Outcomes

As we have seen, prospect theory differs from utility theory in the rel Bmunq rel Bmuationship it suggests between probability and decision weight. Inutility theory, decision weights and probabilities are the same. Thedecision weight of a sure thing is 100, and the weight that corresponds toa 90% chance is exactly 90, which is 9 times more than the decisionweight for a 10% chance. In prospect theory, variations of probability haveless effect on decision weights. An experiment that I mentioned earlier

(PDF) Thinking, fast and slow by daniel kahneman - PDFSLIDE.US (318)

found that the decision weight for a 90% chance was 71.2 and thedecision weight for a 10% chance was 18.6. The ratio of the probabilitieswas 9.0, but the ratio of the decision weights was only 3.83, indicatinginsufficient sensitivity to probability in that range. In both theories, thedecision weights depend only on probability, not on the outcome. Boththeories predict that the decision weight for a 90% chance is the same forwinning $100, receiving a dozen roses, or getting an electric shock. Thistheoretical prediction turns out to be wrong.

Psychologists at the University of Chicago published an article with theattractive title “Money, Kisses, and Electric Shocks: On the AffectivePsychology of Risk.” Their finding was that the valuation of gambles wasmuch less sensitive to probability when the (fictitious) outcomes wereemotional (“meeting and kissing your favorite movie star” or “getting apainful, but not dangerous, electric shock”) than when the outcomes weregains or losses of cash. This was not an isolated finding. Otherresearchers had found, using physiological measures such as heart rate,that the fear of an impending electric shock was essentially uncorrelatedwith the probability of receiving the shock. The mere possibility of a shocktriggered the full-blown fear response. The Chicago team proposed that“affect-laden imagery” overwhelmed the response to probability. Ten yearslater, a team of psychologists at Princeton challenged that conclusion.

The Princeton team argued that the low sensitivity to probability that hadbeen observed for emotional outcomes is normal. Gambles on money arethe exception. The sensitivity to probability is relatively high for thesegambles, because they have a definite expected value.

What amount of cash is as attractive as each of these gambles?

A. 84% chance to win $59B. 84% chance to receive one dozen red roses in a glass vase

What do you notice? The salient difference is that question A is mucheasier than question B. You did not stop to compute the expected value ofthe bet, but you probably knew quickly that it is not far from $50 (in fact it is$49.56), and the vague estimate was sufficient to provide a helpful anchoras you searched for an equally attractive cash gift. No such anchor isavailable for question B, which is therefore much harder to answer.Respondents also assessed the cash equivalent of gambles with a 21%chance to win the two outcomes. As expected, the difference between thehigh-probability and low-probability gambles was much more pronouncedfor the money than for the roses.

(PDF) Thinking, fast and slow by daniel kahneman - PDFSLIDE.US (319)

To bolster their argument that insensitivity to probability is not caused byemotion, the Princeton team compared willingness to pay to avoidgambles:

21% chance (or 84% chance) to spend a weekend paintingsomeone’s three-bedroom apartment

21% chance (or 84% chance) to clean three stalls in a dormitorybath Bmun qbath Bmuroom after a weekend of use

The second outcome is surely much more emotional than the first, but thedecision weights for the two outcomes did not differ. Evidently, the intensityof emotion is not the answer.

Another experiment yielded a surprising result. The participantsreceived explicit price information along with the verbal description of theprize. An example could be:

84% chance to win: A dozen red roses in a glass vase. Value$59.

21% chance to win: A dozen red roses in a glass vase. Value$59.

It is easy to assess the expected monetary value of these gambles, butadding a specific monetary value did not alter the results: evaluationsremained insensitive to probability even in that condition. People whothought of the gift as a chance to get roses did not use price information asan anchor in evaluating the gamble. As scientists sometimes say, this is asurprising finding that is trying to tell us something. What story is it trying totell us?

The story, I believe, is that a rich and vivid representation of theoutcome, whether or not it is emotional, reduces the role of probability inthe evaluation of an uncertain prospect. This hypothesis suggests aprediction, in which I have reasonably high confidence: adding irrelevantbut vivid details to a monetary outcome also disrupts calculation. Compareyour cash equivalents for the following outcomes:

21% (or 84%) chance to receive $59 next Monday

21% (or 84%) chance to receive a large blue cardboard

(PDF) Thinking, fast and slow by daniel kahneman - PDFSLIDE.US (320)

21% (or 84%) chance to receive a large blue cardboardenvelope containing $59 next Monday morning

The new hypothesis is that there will be less sensitivity to probability in thesecond case, because the blue envelope evokes a richer and more fluentrepresentation than the abstract notion of a sum of money. You constructedthe event in your mind, and the vivid image of the outcome exists thereeven if you know that its probability is low. Cognitive ease contributes tothe certainty effect as well: when you hold a vivid image of an event, thepossibility of its not occurring is also represented vividly, andoverweighted. The combination of an enhanced possibility effect with anenhanced certainty effect leaves little room for decision weights to changebetween chances of 21% and 84%.

Vivid Probabilities

The idea that fluency, vividness, and the ease of imagining contribute todecision weights gains support from many other observations. Participantsin a well-known experiment are given a choice of drawing a marble fromone of two urns, in which red marbles win a prize:

Urn A contains 10 marbles, of which 1 is red.Urn B contains 100 marbles, of which 8 are red.

Which urn would you choose? The chances of winning are 10% in urn Aand 8% in urn B, so making the right choice should be easy, but it is not:about 30%–40% of students choose the urn Bmun q urn Bmu with thelarger number of winning marbles, rather than the urn that provides a betterchance of winning. Seymour Epstein has argued that the results illustratethe superficial processing characteristic of System 1 (which he calls theexperiential system).

As you might expect, the remarkably foolish choices that people make inthis situation have attracted the attention of many researchers. The biashas been given several names; following Paul Slovic I will call itdenominator neglect. If your attention is drawn to the winning marbles, youdo not assess the number of nonwinning marbles with the same care. Vividimagery contributes to denominator neglect, at least as I experience it.When I think of the small urn, I see a single red marble on a vaguelydefined background of white marbles. When I think of the larger urn, I seeeight winning red marbles on an indistinct background of white marbles,which creates a more hopeful feeling. The distinctive vividness of thewinning marbles increases the decision weight of that event, enhancing the

(PDF) Thinking, fast and slow by daniel kahneman - PDFSLIDE.US (321)

possibility effect. Of course, the same will be true of the certainty effect. If Ihave a 90% chance of winning a prize, the event of not winning will bemore salient if 10 of 100 marbles are “losers” than if 1 of 10 marbles yieldsthe same outcome.

The idea of denominator neglect helps explain why different ways ofcommunicating risks vary so much in their effects. You read that “a vaccinethat protects children from a fatal disease carries a 0.001% risk ofpermanent disability.” The risk appears small. Now consider anotherdescription of the same risk: “One of 100,000 vaccinated children will bepermanently disabled.” The second statement does something to yourmind that the first does not: it calls up the image of an individual child whois permanently disabled by a vaccine; the 999,999 safely vaccinatedchildren have faded into the background. As predicted by denominatorneglect, low-probability events are much more heavily weighted whendescribed in terms of relative frequencies (how many) than when stated inmore abstract terms of “chances,” “risk,” or “probability” (how likely). As wehave seen, System 1 is much better at dealing with individuals thancategories.

The effect of the frequency format is large. In one study, people who sawinformation about “a disease that kills 1,286 people out of every 10,000”judged it as more dangerous than people who were told about “a diseasethat kills 24.14% of the population.” The first disease appears morethreatening than the second, although the former risk is only half as largeas the latter! In an even more direct demonstration of denominator neglect,“a disease that kills 1,286 people out of every 10,000” was judged moredangerous than a disease that “kills 24.4 out of 100.” The effect wouldsurely be reduced or eliminated if participants were asked for a directcomparison of the two formulations, a task that explicitly calls for System 2.Life, however, is usually a between-subjects experiment, in which you seeonly one formulation at a time. It would take an exceptionally active System2 to generate alternative formulations of the one you see and to discoverthat they evoke a different response.

Experienced forensic psychologists and psychiatrists are not immune tothe effects of the format in which risks are expressed. In one experiment,professionals evaluated whether it was safe to discharge from thepsychiatric hospital a patient, Mr. Jones, with a history of violence. Theinformation they received included an expert’s assessment of the risk. Thesame statistics were described in two ways:

Patients similar to Mr. Jones are estimated to have a 10%probability of committing an act of violence against others duringthe first several months after discharge.

(PDF) Thinking, fast and slow by daniel kahneman - PDFSLIDE.US (322)

Of every 100 patients similar to Mr. Jones, 10 are estimated tocommit an act of violence against others during the first severalmonths after discharge.

The professionals who saw the frequency format were almost twice aslikely to deny the discharge (41%, compared to 21% in the probabilityformat). The more vivid description produces a higher decision weight forthe same probability.

The power of format creates opportunities for manipulation, whichpeople with an axe to grind know how to exploit. Slovic and his colleaguescite an article that states that “approximately 1,000 homicides a year arecommitted nationwide by seriously mentally ill individuals who are nottaking their medication.” Another way of expressing the same fact is that“1,000 out of 273,000,000 Americans will die in this manner each year.”Another is that “the annual likelihood of being killed by such an individual isapproximately 0.00036%.” Still another: “1,000 Americans will die in thismanner each year, or less than one-thirtieth the number who will die ofsuicide and about one-fourth the number who will die of laryngeal cancer.”Slovic points out that “these advocates are quite open about theirmotivation: they want to frighten the general public about violence bypeople with mental disorder, in the hope that this fear will translate intoincreased funding for mental health services.”

A good attorney who wishes to cast doubt on DNA evidence will not tellthe jury that “the chance of a false match is 0.1%.” The statement that “afalse match occurs in 1 of 1,000 capital cases” is far more likely to passthe threshold of reasonable doubt. The jurors hearing those words areinvited to generate the image of the man who sits before them in thecourtroom being wrongly convicted because of flawed DNA evidence. Theprosecutor, of course, will favor the more abstract frame—hoping to fill thejurors’ minds with decimal points.

Decisions from Global Impressions

The evidence suggests the hypothesis that focal attention and saliencecontribute to both the overestimation of unlikely events and theoverweighting of unlikely outcomes. Salience is enhanced by meremention of an event, by its vividness, and by the format in which probabilityis described. There are exceptions, of course, in which focusing on anevent does not raise its probability: cases in which an erroneous theorymakes an event appear impossible even when you think about it, or cases

(PDF) Thinking, fast and slow by daniel kahneman - PDFSLIDE.US (323)

makes an event appear impossible even when you think about it, or casesin which an inability to imagine how an outcome might come about leavesyou convinced that it will not happen. The bias toward overestimation andoverweighting of salient events is not an absolute rule, but it is large androbust.

There has been much interest in recent years in studies of choice fromexperience, which follow different rules from the choices from descriptionthat are analyzed in prospect theory. Participants in a typical experimentface two buttons. When pressed, each button produces either a monetaryreward or nothing, and the outcome is drawn randomly according to thespecifications of a prospect (for example, “5% to win $12” or “95% chanceto win $1”). The process is truly random, s Bmun qm, s Bmuo there is noguarantee that the sample a participant sees exactly represents thestatistical setup. The expected values associated with the two buttons areapproximately equal, but one is riskier (more variable) than the other. (Forexample, one button may produce $10 on 5% of the trials and the other $1on 50% of the trials). Choice from experience is implemented by exposingthe participant to many trials in which she can observe the consequencesof pressing one button or another. On the critical trial, she chooses one ofthe two buttons, and she earns the outcome on that trial. Choice fromdescription is realized by showing the subject the verbal description of therisky prospect associated with each button (such as “5% to win $12”) andasking her to choose one. As expected from prospect theory, choice fromdescription yields a possibility effect—rare outcomes are overweightedrelative to their probability. In sharp contrast, overweighting is neverobserved in choice from experience, and underweighting is common.

The experimental situation of choice by experience is intended torepresent many situations in which we are exposed to variable outcomesfrom the same source. A restaurant that is usually good may occasionallyserve a brilliant or an awful meal. Your friend is usually good company, buthe sometimes turns moody and aggressive. California is prone toearthquakes, but they happen rarely. The results of many experimentssuggest that rare events are not overweighted when we make decisionssuch as choosing a restaurant or tying down the boiler to reduceearthquake damage.

The interpretation of choice from experience is not yet settled, but thereis general agreement on one major cause of underweighting of rareevents, both in experiments and in the real world: many participants neverexperience the rare event! Most Californians have never experienced amajor earthquake, and in 2007 no banker had personally experienced adevastating financial crisis. Ralph Hertwig and Ido Erev note that “chancesof rare events (such as the burst of housing bubbles) receive less impact

(PDF) Thinking, fast and slow by daniel kahneman - PDFSLIDE.US (324)

than they deserve according to their objective probabilities.” They point tothe public’s tepid response to long-term environmental threats as anexample.

These examples of neglect are both important and easily explained, butunderweighting also occurs when people have actually experienced therare event. Suppose you have a complicated question that two colleagueson your floor could probably answer. You have known them both for yearsand have had many occasions to observe and experience their character.Adele is fairly consistent and generally helpful, though not exceptional onthat dimension. Brian is not quite as friendly and helpful as Adele most ofthe time, but on some occasions he has been extremely generous with histime and advice. Whom will you approach?

Consider two possible views of this decision:

It is a choice between two gambles. Adele is closer to a sure thing;the prospect of Brian is more likely to yield a slightly inferioroutcome, with a low probability of a very good one. The rare eventwill be overweighted by a possibility effect, favoring Brian.It is a choice between your global impressions of Adele and Brian.The good and the bad experiences you have had are pooled in yourrepresentation of their normal behavior. Unless the rare event is soextreme that it comes to mind separately (Brian once verballyabused a colleague who asked for his help), the norm will be biasedtoward typical and recent instances, favoring Adele.

In a two-system mind, the second interpretation a Bmun qon a Bmuppearsfar more plausible. System 1 generates global representations of Adeleand Brian, which include an emotional attitude and a tendency to approachor avoid. Nothing beyond a comparison of these tendencies is needed todetermine the door on which you will knock. Unless the rare event comesto your mind explicitly, it will not be overweighted. Applying the same ideato the experiments on choice from experience is straightforward. As theyare observed generating outcomes over time, the two buttons developintegrated “personalities” to which emotional responses are attached.

The conditions under which rare events are ignored or overweighted arebetter understood now than they were when prospect theory wasformulated. The probability of a rare event will (often, not always) beoverestimated, because of the confirmatory bias of memory. Thinkingabout that event, you try to make it true in your mind. A rare event will be

(PDF) Thinking, fast and slow by daniel kahneman - PDFSLIDE.US (325)

overweighted if it specifically attracts attention. Separate attention iseffectively guaranteed when prospects are described explicitly (“99%chance to win $1,000, and 1% chance to win nothing”). Obsessiveconcerns (the bus in Jerusalem), vivid images (the roses), concreterepresentations (1 of 1,000), and explicit reminders (as in choice fromdescription) all contribute to overweighting. And when there is nooverweighting, there will be neglect. When it comes to rare probabilities,our mind is not designed to get things quite right. For the residents of aplanet that may be exposed to events no one has yet experienced, this isnot good news.

Speaking of Rare Events

“Tsunamis are very rare even in Japan, but the image is so vividand compelling that tourists are bound to overestimate theirprobability.”

“It’s the familiar disaster cycle. Begin by exaggeration andoverweighting, then neglect sets in.”

“We shouldn’t focus on a single scenario, or we will overestimateits probability. Let’s set up specific alternatives and make theprobabilities add up to 100%.”

“They want people to be worried by the risk. That’s why theydescribe it as 1 death per 1,000. They’re counting ondenominator neglect.”

(PDF) Thinking, fast and slow by daniel kahneman - PDFSLIDE.US (326)

Risk Policies

Imagine that you face the following pair of concurrent decisions. Firstexamine both decisions, then make your choices.

Decision (i): Choose between

A. sure gain of $240B. 25% chance to gain $1,000 and 75% chance to gain nothing

Decision (ii): Choose between

C. sure loss of $750D. 75% chance to lose $1,000 and 25% chance to lose nothing

This pair of choice problems has an important place in the history ofprospect theory, and it has new things to tell us about rationality. As youskimmed the two problems, your initial reaction to the sure things (A andC) was attraction to the first and aversion to the second. The emotionalevaluation of “sure gain” and “sure loss” is an automatic reaction of System1, which certainly occurs before the more effortful (and optional)computation of the expected values of the two gambles (respectively, again of $250 and a loss of $750). Most people’s choices correspond to thepredilections of System 1, and large majorities prefer A to B and D to C.As in many other choices that involve moderate or high probabilities,people tend to be risk averse in the domain of gains and risk seeking inthe domain of losses. In the original experiment that Amos and I carriedout, 73% of respondents chose A in decision i and D in decision ii andonly 3% favored the combination of B and C.

You were asked to examine both options before making your firstchoice, and you probably did so. But one thing you surely did not do: youdid not compute the possible results of the four combinations of choices (Aand C, A and D, B and C, B and D) to determine which combination youlike best. Your separate preferences for the two problems were intuitivelycompelling and there was no reason to expect that they could lead totrouble. Furthermore, combining the two decision problems is a laboriousexercise that you would need paper and pencil to complete. You did not doit. Now consider the following choice problem:

(PDF) Thinking, fast and slow by daniel kahneman - PDFSLIDE.US (327)

AD. 25% chance to win $240 and 75% chance to lose $760BC. 25% chance to win $250 and 75% chance to lose $750

This choice is easy! Option BC actually dominates option AD (thetechnical term for one option being unequivocally better than another). Youalready know what comes next. The dominant option in AD is thecombination of the two rejected options in the first pair of decisionproblems, the one that only 3% of respondents favored in our originalstudy. The inferior option BC was preferred by 73% of respondents.

Broad or Narrow?

This set of choices has a lot to tell us about the limits of human rationality.For one thing, it helps us see the logical consistency of Humanpreferences for what it is—a hopeless mirage. Have another look at thelast problem, the easy one. Would you have imagined the possibility ofdecomposing this obvious choice problem into a pair of problems thatwould lead a large majority of people to choose an inferior option? This isgenerally true: every simple choice formulated in terms of gains and lossescan be deconstructed in innumerable ways into a combination of choices,yielding preferences that are likely to be inconsistent.

The example also shows that it is costly to be risk averse for gains andrisk seeking for losses. These attitudes make you willing to pay a premiumto obtain a sure gain rather than face a gamble, and also willing to pay apremium (in expected value) to avoid a sure loss. Both payments come outof the same pocket, and when you face both kinds of problems at once, thediscrepant attitudes are unlikely to be optimal.

There were tw Bght hecome oo ways of construing decisions i and ii:

narrow framing: a sequence of two simple decisions, consideredseparatelybroad framing: a single comprehensive decision, with four options

Broad framing was obviously superior in this case. Indeed, it will besuperior (or at least not inferior) in every case in which several decisionsare to be contemplated together. Imagine a longer list of 5 simple (binary)decisions to be considered simultaneously. The broad (comprehensive)frame consists of a single choice with 32 options. Narrow framing will yielda sequence of 5 simple choices. The sequence of 5 choices will be one of

(PDF) Thinking, fast and slow by daniel kahneman - PDFSLIDE.US (328)

the 32 options of the broad frame. Will it be the best? Perhaps, but not verylikely. A rational agent will of course engage in broad framing, but Humansare by nature narrow framers.

The ideal of logical consistency, as this example shows, is notachievable by our limited mind. Because we are susceptible to WY SIATIand averse to mental effort, we tend to make decisions as problems arise,even when we are specifically instructed to consider them jointly. We haveneither the inclination nor the mental resources to enforce consistency onour preferences, and our preferences are not magically set to be coherent,as they are in the rational-agent model.

Samuelson’s Problem

The great Paul Samuelson—a giant among the economists of thetwentieth century—famously asked a friend whether he would accept agamble on the toss of a coin in which he could lose $100 or win $200. Hisfriend responded, “I won’t bet because I would feel the $100 loss morethan the $200 gain. But I’ll take you on if you promise to let me make 100such bets.” Unless you are a decision theorist, you probably share theintuition of Samuelson’s friend, that playing a very favorable but riskygamble multiple times reduces the subjective risk. Samuelson found hisfriend’s answer interesting and went on to analyze it. He proved that undersome very specific conditions, a utility maximizer who rejects a singlegamble should also reject the offer of many.

Remarkably, Samuelson did not seem to mind the fact that his proof,which is of course valid, led to a conclusion that violates common sense, ifnot rationality: the offer of a hundred gambles is so attractive that no saneperson would reject it. Matthew Rabin and Richard Thaler pointed out that“the aggregated gamble of one hundred 50–50 lose $100/gain $200 betshas an expected return of $5,000, with only a 1/2,300 chance of losing anymoney and merely a 1/62,000 chance of losing more than $1,000.” Theirpoint, of course, is that if utility theory can be consistent with such a foolishpreference under any circumstances, then something must be wrong with itas a model of rational choice. Samuelson had not seen Rabin’s proof ofthe absurd consequences of severe loss aversion for small bets, but hewould surely not have been surprised by it. His willingness even toconsider the possibility that it could be rational to reject the packagetestifies to the powerful hold of the rational model.

Let us assume that a very simple value function describes thepreferences of Samuelson’s friend (call him Sam). To express his aversionto losses Sam first rewrites the bet, after multiplying each loss by a factor

(PDF) Thinking, fast and slow by daniel kahneman - PDFSLIDE.US (329)

of 2. He then computes the expected value of the rewritten bet. Here arethe results, for one, two, or three tosses. They are sufficiently instructive todeserve some Bght iciof 2

You can see in the display that the gamble has an expected value of 50.However, one toss is worth nothing to Sam because he feels that the painof losing a dollar is twice as intense as the pleasure of winning a dollar.After rewriting the gamble to reflect his loss aversion, Sam will find that thevalue of the gamble is 0.

Now consider two tosses. The chances of losing have gone down to25%. The two extreme outcomes (lose 200 or win 400) cancel out in value;they are equally likely, and the losses are weighted twice as much as thegain. But the intermediate outcome (one loss, one gain) is positive, and sois the compound gamble as a whole. Now you can see the cost of narrowframing and the magic of aggregating gambles. Here are two favorablegambles, which individually are worth nothing to Sam. If he encounters theoffer on two separate occasions, he will turn it down both times. However,if he bundles the two offers together, they are jointly worth $50!

Things get even better when three gambles are bundled. The extremeoutcomes still cancel out, but they have become less significant. The thirdtoss, although worthless if evaluated on its own, has added $62.50 to thetotal value of the package. By the time Sam is offered five gambles, theexpected value of the offer will be $250, his probability of losing anythingwill be 18.75%, and his cash equivalent will be $203.125. The notableaspect of this story is that Sam never wavers in his aversion to losses.However, the aggregation of favorable gambles rapidly reduces the

(PDF) Thinking, fast and slow by daniel kahneman - PDFSLIDE.US (330)

probability of losing, and the impact of loss aversion on his preferencesdiminishes accordingly.

Now I have a sermon ready for Sam if he rejects the offer of a singlehighly favorable gamble played once, and for you if you share hisunreasonable aversion to losses:

I sympathize with your aversion to losing any gamble, but it iscosting you a lot of money. Please consider this question: Areyou on your deathbed? Is this the last offer of a small favorablegamble that you will ever consider? Of course, you are unlikely tobe offered exactly this gamble again, but you will have manyopportunities to consider attractive gambles with stakes that arevery small relative to your wealth. You will do yourself a largefinancial favor if you are able to see each of these gambles aspart of a bundle of small gambles and rehearse the mantra thatwill get you significantly closer to economic rationality: you win afew, you lose a few. The main purpose of the mantra is to controlyour emotional response when you do lose. If you can trust it to beeffective, you should remind yourself of it when deciding whetheror not to accept a small risk with positive expected value.Remember these qualifications when using the mantra:

It works when the gambles are genuinely independent of each other;it does not apply to multiple investments in the same industry, whichwould all go bad together.It works only when the possible loss does not cause you to worryabout your total wealth. If you would take the loss as significant badnews about your economic future, watch it!It should not be applied to long shots, where the probability ofwinning is very small for each bet.

If you have the emotional discipline that this rule requires, Bght l dfor e you will never consider a small gamble in isolation or be lossaverse for a small gamble until you are actually on your deathbed—and not even then.

This advice is not impossible to follow. Experienced traders in financial

(PDF) Thinking, fast and slow by daniel kahneman - PDFSLIDE.US (331)

markets live by it every day, shielding themselves from the pain of lossesb y broad framing. As was mentioned earlier, we now know thatexperimental subjects could be almost cured of their loss aversion (in aparticular context) by inducing them to “think like a trader,” just asexperienced baseball card traders are not as susceptible to theendowment effect as novices are. Students made risky decisions (toaccept or reject gambles in which they could lose) under differentinstructions. In the narrow-framing condition, they were told to “make eachdecision as if it were the only one” and to accept their emotions. Theinstructions for broad framing of a decision included the phrases “imagineyourself as a trader,” “you do this all the time,” and “treat it as one of manymonetary decisions, which will sum together to produce a ‘portfolio.’” Theexperimenters assessed the subjects’ emotional response to gains andlosses by physiological measures, including changes in the electricalconductance of the skin that are used in lie detection. As expected, broadframing blunted the emotional reaction to losses and increased thewillingness to take risks.

The combination of loss aversion and narrow framing is a costly curse.Individual investors can avoid that curse, achieving the emotional benefitsof broad framing while also saving time and agony, by reducing thefrequency with which they check how well their investments are doing.Closely following daily fluctuations is a losing proposition, because thepain of the frequent small losses exceeds the pleasure of the equallyfrequent small gains. Once a quarter is enough, and may be more thanenough for individual investors. In addition to improving the emotionalquality of life, the deliberate avoidance of exposure to short-term outcomesimproves the quality of both decisions and outcomes. The typical short-term reaction to bad news is increased loss aversion. Investors who getaggregated feedback receive such news much less often and are likely tobe less risk averse and to end up richer. You are also less prone touseless churning of your portfolio if you don’t know how every stock in it isdoing every day (or every week or even every month). A commitment not tochange one’s position for several periods (the equivalent of “locking in” aninvestment) improves financial performance.

Risk Policies

Decision makers who are prone to narrow framing construct a preferenceevery time they face a risky choice. They would do better by having a riskpolicy that they routinely apply whenever a relevant problem arises.Familiar examples of risk policies are “always take the highest possible

(PDF) Thinking, fast and slow by daniel kahneman - PDFSLIDE.US (332)

deductible when purchasing insurance” and “never buy extendedwarranties.” A risk policy is a broad frame. In the insurance examples, youexpect the occasional loss of the entire deductible, or the occasionalfailure of an uninsured product. The relevant issue is your ability to reduceor eliminate the pain of the occasional loss by the thought that the policythat left you exposed to it will almost certainly be financially advantageousover the long run.

A risk policy that aggregates decisions is analogous to the outside viewof planning problems that I discussed earlier. The outside view shift s thefocus from the specifics of the current situation to Bght pecicy tthestatistics of outcomes in similar situations. The outside view is a broadframe for thinking about plans. A risk policy is a broad frame that embedsa particular risky choice in a set of similar choices.

The outside view and the risk policy are remedies against two distinctbiases that affect many decisions: the exaggerated optimism of theplanning fallacy and the exaggerated caution induced by loss aversion.The two biases oppose each other. Exaggerated optimism protectsindividuals and organizations from the paralyzing effects of loss aversion;loss aversion protects them from the follies of overconfident optimism. Theupshot is rather comfortable for the decision maker. Optimists believe thatthe decisions they make are more prudent than they really are, and loss-averse decision makers correctly reject marginal propositions that theymight otherwise accept. There is no guarantee, of course, that the biasescancel out in every situation. An organization that could eliminate bothexcessive optimism and excessive loss aversion should do so. Thecombination of the outside view with a risk policy should be the goal.

Richard Thaler tells of a discussion about decision making he had withthe top managers of the 25 divisions of a large company. He asked themto consider a risky option in which, with equal probabilities, they could losea large amount of the capital they controlled or earn double that amount.None of the executives was willing to take such a dangerous gamble.Thaler then turned to the CEO of the company, who was also present, andasked for his opinion. Without hesitation, the CEO answered, “I would likeall of them to accept their risks.” In the context of that conversation, it wasnatural for the CEO to adopt a broad frame that encompassed all 25 bets.Like Sam facing 100 coin tosses, he could count on statistical aggregationto mitigate the overall risk.

Speaking of Risk Policies

“Tell her to think like a trader! You win a few, you lose a few.”

(PDF) Thinking, fast and slow by daniel kahneman - PDFSLIDE.US (333)

“I decided to evaluate my portfolio only once a quarter. I am tooloss averse to make sensible decisions in the face of daily pricefluctuations.”

“They never buy extended warranties. That’s their risk policy.”

“Each of our executives is loss averse in his or her domain.That’s perfectly natural, but the result is that the organization is nottaking enough risk.”

(PDF) Thinking, fast and slow by daniel kahneman - PDFSLIDE.US (334)

Keeping Score

Except for the very poor, for whom income coincides with survival, the mainmotivators of money-seeking are not necessarily economic. For thebillionaire looking for the extra billion, and indeed for the participant in anexperimental economics project looking for the extra dollar, money is aproxy for points on a scale of self-regard and achievement. These rewardsand punishments, promises and threats, are all in our heads. We carefullykeep score of them. They shape o C Th5ur preferences and motivate ouractions, like the incentives provided in the social environment. As a result,we refuse to cut losses when doing so would admit failure, we are biasedagainst actions that could lead to regret, and we draw an illusory but sharpdistinction between omission and commission, not doing and doing,because the sense of responsibility is greater for one than for the other.The ultimate currency that rewards or punishes is often emotional, a formof mental self-dealing that inevitably creates conflicts of interest when theindividual acts as an agent on behalf of an organization.

Mental Accounts

Richard Thaler has been fascinated for many years by analogies betweenthe world of accounting and the mental accounts that we use to organizeand run our lives, with results that are sometimes foolish and sometimesvery helpful. Mental accounts come in several varieties. We hold our moneyin different accounts, which are sometimes physical, sometimes onlymental. We have spending money, general savings, earmarked savings forour children’s education or for medical emergencies. There is a clearhierarchy in our willingness to draw on these accounts to cover currentneeds. We use accounts for self-control purposes, as in making ahousehold budget, limiting the daily consumption of espressos, orincreasing the time spent exercising. Often we pay for self-control, forinstance simultaneously putting money in a savings account andmaintaining debt on credit cards. The Econs of the rational-agent modeldo not resort to mental accounting: they have a comprehensive view ofoutcomes and are driven by external incentives. For Humans, mentalaccounts are a form of narrow framing; they keep things under control andmanageable by a finite mind.

Mental accounts are used extensively to keep score. Recall thatprofessional golfers putt more successfully when working to avoid a bogeythan to achieve a birdie. One conclusion we can draw is that the bestgolfers create a separate account for each hole; they do not only maintain

(PDF) Thinking, fast and slow by daniel kahneman - PDFSLIDE.US (335)

a single account for their overall success. An ironic example that Thalerrelated in an early article remains one of the best illustrations of howmental accounting affects behavior:

Two avid sports fans plan to travel 40 miles to see a basketballgame. One of them paid for his ticket; the other was on his way topurchase a ticket when he got one free from a friend. A blizzard isannounced for the night of the game. Which of the two ticketholders is more likely to brave the blizzard to see the game?

The answer is immediate: we know that the fan who paid for his ticket ismore likely to drive. Mental accounting provides the explanation. Weassume that both fans set up an account for the game they hoped to see.Missing the game will close the accounts with a negative balance.Regardless of how they came by their ticket, both will be disappointed—but the closing balance is distinctly more negative for the one who bought aticket and is now out of pocket as well as deprived of the game. Becausestaying home is worse for this individual, he is more motivated to see thegame and therefore more likely to make the attempt to drive into a blizzard.These are tacit calculations of emotional balance, of the kind that System 1performs without deliberation. The emotions that people attach to the stateof their mental accounts are not acknowledged in standard economictheory. An Econ would realize that the ticket has already been paid for andcannot be returned. Its cost is “sunk” and the Econ would not care whetherhe had bought the ticket to the game or got it from a friend (if Eco BTh5motketns have friends). To implement this rational behavior, System 2would have to be aware of the counterfactual possibility: “Would I still driveinto this snowstorm if I had gotten the ticket free from a friend?” It takes anactive and disciplined mind to raise such a difficult question.

A related mistake afflicts individual investors when they sell stocks fromtheir portfolio:

You need money to cover the costs of your daughter’s weddingand will have to sell some stock. You remember the price atwhich you bought each stock and can identify it as a “winner,”currently worth more than you paid for it, or as a loser. Among thestocks you own, Blueberry Tiles is a winner; if you sell it today youwill have achieved a gain of $5,000. You hold an equalinvestment in Tiffany Motors, which is currently worth $5,000 lessthan you paid for it. The value of both stocks has been stable inrecent weeks. Which are you more likely to sell?

(PDF) Thinking, fast and slow by daniel kahneman - PDFSLIDE.US (336)

A plausible way to formulate the choice is this: “I could close the BlueberryTiles account and score a success for my record as an investor.Alternatively, I could close the Tiffany Motors account and add a failure tomy record. Which would I rather do?” If the problem is framed as a choicebetween giving yourself pleasure and causing yourself pain, you willcertainly sell Blueberry Tiles and enjoy your investment prowess. As mightbe expected, finance research has documented a massive preference forselling winners rather than losers—a bias that has been given an opaquelabel: the disposition effect.

The disposition effect is an instance of narrow framing. The investor hasset up an account for each share that she bought, and she wants to closeevery account as a gain. A rational agent would have a comprehensiveview of the portfolio and sell the stock that is least likely to do well in thefuture, without considering whether it is a winner or a loser. Amos told meof a conversation with a financial adviser, who asked him for a completelist of the stocks in his portfolio, including the price at which each had beenpurchased. When Amos asked mildly, “Isn’t it supposed not to matter?” theadviser looked astonished. He had apparently always believed that thestate of the mental account was a valid consideration.

Amos’s guess about the financial adviser’s beliefs was probably right,but he was wrong to dismiss the buying price as irrelevant. The purchaseprice does matter and should be considered, even by Econs. Thedisposition effect is a costly bias because the question of whether to sellwinners or losers has a clear answer, and it is not that it makes nodifference. If you care about your wealth rather than your immediateemotions, you will sell the loser Tiffany Motors and hang on to the winningBlueberry Tiles. At least in the United States, taxes provide a strongincentive: realizing losses reduces your taxes, while selling winnersexposes you to taxes. This elementary fact of financial life is actually knownto all American investors, and it determines the decisions they makeduring one month of the year—investors sell more losers in December,when taxes are on their mind. The tax advantage is available all year, ofcourse, but for 11 months of the year mental accounting prevails overfinancial common sense. Another argument against selling winners is thewell-documented market anomaly that stocks that recently gained in valueare likely to go on gaining at least for a short while. The net effect is large:the expected after-tax extra return of selling Tiffany rather than Blueberry is3.4% over the next year. Cl B Th5inge liosing a mental account with a gainis a pleasure, but it is a pleasure you pay for. The mistake is not one thatan Econ would ever make, and experienced investors, who are using theirSystem 2, are less susceptible to it than are novices.

(PDF) Thinking, fast and slow by daniel kahneman - PDFSLIDE.US (337)

A rational decision maker is interested only in the future consequencesof current investments. Justifying earlier mistakes is not among the Econ’sconcerns. The decision to invest additional resources in a losing account,when better investments are available, is known as the sunk-cost fallacy, acostly mistake that is observed in decisions large and small. Driving intothe blizzard because one paid for tickets is a sunk-cost error.

Imagine a company that has already spent $50 million on a project. Theproject is now behind schedule and the forecasts of its ultimate returns areless favorable than at the initial planning stage. An additional investment of$60 million is required to give the project a chance. An alternative proposalis to invest the same amount in a new project that currently looks likely tobring higher returns. What will the company do? All too often a companyafflicted by sunk costs drives into the blizzard, throwing good money afterbad rather than accepting the humiliation of closing the account of a costlyfailure. This situation is in the top-right cell of the fourfold pattern, where thechoice is between a sure loss and an unfavorable gamble, which is oftenunwisely preferred.

The escalation of commitment to failing endeavors is a mistake from theperspective of the firm but not necessarily from the perspective of theexecutive who “owns” a floundering project. Canceling the project will leavea permanent stain on the executive’s record, and his personal interests areperhaps best served by gambling further with the organization’s resourcesin the hope of recouping the original investment—or at least in an attemptto postpone the day of reckoning. In the presence of sunk costs, themanager’s incentives are misaligned with the objectives of the firm and itsshareholders, a familiar type of what is known as the agency problem.Boards of directors are well aware of these conflicts and often replace aCEO who is encumbered by prior decisions and reluctant to cut losses.The members of the board do not necessarily believe that the new CEO ismore competent than the one she replaces. They do know that she doesnot carry the same mental accounts and is therefore better able to ignorethe sunk costs of past investments in evaluating current opportunities.

The sunk-cost fallacy keeps people for too long in poor jobs, unhappymarriages, and unpromising research projects. I have often observedyoung scientists struggling to salvage a doomed project when they wouldbe better advised to drop it and start a new one. Fortunately, researchsuggests that at least in some contexts the fallacy can be overcome. Thesunk-cost fallacy is identified and taught as a mistake in both economicsand business courses, apparently to good effect: there is evidence thatgraduate students in these fields are more willing than others to walk awayfrom a failing project.