Two psychologists proved smart judgment fails in predictable ways

By Books (/journals/books)

In 60 seconds Daniel Kahneman and Amos Tversky’s collaboration showed human judgment runs on mental shortcuts, not careful calculation. Representativeness, availability, and anchoring make even trained experts ignore base rates, overreact to vivid cases, and cling to arbitrary numbers. Prospect theory proved people weigh losses roughly twice as heavily as gains and flip preferences when the same numbers are framed differently. The same biases warped basketball scouting, medical diagnosis, and intelligence forecasts until organizations checked intuition against data. Their partnership itself collapsed under unequal credit and fame, illustrating the attribution errors they had documented. Tversky died in 1996; the 2002 Nobel went to Kahneman alone for work neither could have done solo.

The big idea

Michael Lewis’s 2016 narrative The Undoing Project shows that Daniel Kahneman and Amos Tversky overturned the assumption of rational human judgment by proving people rely on predictable heuristics and feel losses far more than gains. Their decades-long conversation produced prospect theory and the biases that reshaped economics, medicine, and sports, yet the same need for credit that they studied eventually fractured the partnership that made the discoveries possible.

The book in brief

Lewis opens with the odd couple at Hebrew University in the late 1960s: Kahneman the anxious reviser, Tversky the supremely confident showman. Colleagues expected little; instead the two discovered that arguing and trading sentences produced ideas neither could reach alone. Their laughter-filled sessions became the engine of a new science of judgment.

Both men had already watched expert confidence fail in the Israeli military. Kahneman saw officer-interview ratings bear almost no relation to later performance and named the gap the illusion of validity. Tversky’s combat experience left him skeptical of tidy stories about bravery or folly. Those lived failures set the question that defined their careers: why do smart, motivated people trust judgment that performs no better than chance?

They answered with simple experiments that exposed stable shortcuts. Representativeness led people, even statistics graduate students, to rate a conjunction as more probable than one of its parts in the Linda problem. Availability made memorable plane crashes and shark attacks feel commoner than quiet statistical killers. Anchoring pulled estimates toward a rigged wheel spin of 10 or 65 even when subjects knew the number was irrelevant. Each heuristic traded accuracy for speed and survived because it worked well enough in ordinary life.

In 1979 they published prospect theory, showing people evaluate outcomes as gains or losses against a reference point and that losses hurt roughly twice as much as equivalent gains please. The Asian disease problem proved identical statistics produce opposite choices when framed as lives saved versus lives lost. The finding dismantled classical expected-utility models and later earned Kahneman a share of the 2002 Nobel in Economics.

Lewis then follows the ideas into basketball front offices, hospital wards, and forecasting shops. Houston Rockets general manager Daryl Morey fought scouts who anchored on how a player looked; physician Donald Redelmeier showed doctors overweighting recent dramatic cases. Expertise, it turned out, often amplified overconfidence rather than curing it. Organizations that forced judgment to face base rates and data improved drafts, diagnoses, and risk calls.

The final movement turns inward. Tversky’s simulation heuristic explained why near misses generate more regret than clear failures and why deviations from routine attract more blame. By the early 1980s unequal public credit cooled the friendship into silence. Tversky died of melanoma in 1996 at fifty-nine; the Nobel six years later could not name him. Kahneman spent the rest of his career insisting the prize belonged to both, leaving Lewis’s double story complete: the mind’s predictable errors, and the human cost of the partnership that revealed them.

Context

Michael Lewis, a narrative journalist rather than a psychologist, published The Undoing Project in 2016. The book mattered because it gave general readers the human story and the rigorous, testable account of how and why people predictably err, carrying the heuristics-and-biases program from psychology journals into economics, medicine, sports management, and policy.

The key ideas

1

Friction creates the ideas

Great intellectual breakthroughs sometimes require the friction of two profoundly different minds rather than the isolated brilliance of one.

DDaniel KahnemanAnxious, self-doubting, constant reviser of his own ideas AAmos TverskyDecisive, supremely confident, famous for being right out loud

Kahneman was anxious, self-doubting, and forever revising his own thoughts. Tversky was decisive, almost never wrong in colleagues’ eyes, and fearless in argument. When they started talking regularly at Hebrew University, people expected the temperaments to cancel each other out.

Instead the clash became generative. They wrote by trading sentences in one room, laughing so hard at their thought experiments that neighbors complained. Neither could later say who had produced which line because the ideas emerged from the exchange itself.

Lewis treats this chemistry as the book’s central claim about scientific creativity: the psychology of collaboration can matter as much as the psychology of the solitary genius. An insecure worrier paired with a confident showman turned out to be exactly the combination that noticed how ordinary people, including experts, systematically misjudge probability and risk.

In practice: They drafted papers by literally passing sentences back and forth until the finished text belonged to neither man alone.

2

War planted the doubt

Direct exposure to life-and-death military decisions gave both men early evidence that confident expert judgment is often no better than a guess.

Interviewers’ feeling Certain they could spot leadership potential in a short meeting

Later record Confidence showed almost no link to actual performance

Before either became a psychologist, both served in the Israeli military and watched real decisions under pressure. Kahneman evaluated candidates for officer training, built structured interviews, then checked the ratings against actual performance.

Interviewers felt certain they could spot leadership in a short conversation; their track record showed almost no link between that certainty and later results. He called the pattern the illusion of validity. Tversky, a decorated paratrooper, once ran into live fire to drag a soldier from a bomb—an act that hardened his skepticism of neat explanations for brave or foolish behavior.

These experiences were not colorful backstory. They planted the career-long question: why do trained, motivated people trust judgment that performs no better than chance? The answer would come from controlled experiments, not philosophy.

In practice: Kahneman found that officers’ confidence in their interview predictions had almost no relationship to how the candidates later performed.

3

Resemblance beats base rates

People estimate probability by how closely something matches a familiar stereotype, not by calculating actual odds, producing predictable errors even in experts.

Logical ranking Bank teller must be at least as likely as bank teller who is also a feminist activist

Typical ranking Subjects rated the activist teller as more probable because she matched the description

Kahneman and Tversky named the shortcut the representativeness heuristic. Asked whether a shy, detail-oriented man is likelier a farmer or a librarian, people ignore that farmers vastly outnumber librarians and answer by stereotype fit.

Their best-known demonstration is the Linda problem: a woman described as socially engaged and politically active, then ranked as more likely “bank teller and active in the feminist movement” than simply “bank teller.” A majority, including statistics graduate students, committed the conjunction fallacy—rating a subset as more probable than the larger set that contains it.

The error is not ignorance. The heuristic runs below conscious reasoning, which is why vivid specific stories feel more probable than dry base rates. That single mechanism explains why stereotypes routinely defeat statistics.

In practice: In the Linda problem most subjects, including trained graduate students, rated the more specific conjunction as likelier than the single broader category.

4

Ease of recall warps risk

People judge how common or dangerous something is by how easily examples come to mind, so memorable events distort risk perception far more than statistics do.

What memory supplies Plane crashes, shark attacks, terrorist incidents

What the numbers show Far rarer than ordinary, under-reported causes of death

The availability heuristic substitutes speed of recall for actual frequency. Subjects estimated that words starting with K are more common than words with K in the third position simply because the former come to mind faster.

Applied to danger, the same shortcut makes plane crashes, shark attacks, and terrorist incidents feel far likelier than their base rates, while common but unreported causes of death barely register. One vivid, widely covered event can shift public fear for years.

Institutions and media therefore warp collective judgment simply by choosing which stories to repeat. A rare dramatic case broadcast endlessly outweighs a mountain of quiet data. The bias is directional, always toward the memorable rather than the frequent, which is why explaining it rarely erases it.

In practice: People treat widely reported plane crashes and shark attacks as major risks while ignoring statistically commoner causes of death that rarely make headlines.

5

Arbitrary numbers pull estimates

An arbitrary number introduced early anchors later judgments even when the person knows the number is meaningless.

Wheel stops at 10 Later guesses about African UN membership ran lower

Wheel stops at 65 Later guesses ran higher even though everyone knew the spin was irrelevant

In a classic experiment Kahneman and Tversky spun a wheel rigged to stop at 10 or 65, then asked for the percentage of African countries in the United Nations. Subjects who saw 65 gave higher guesses than those who saw 10, despite knowing the wheel had no connection to the answer.

Once a figure enters the mind it drags subsequent estimates toward it. The mind almost never starts from a blank slate; every guess is contaminated by whatever number appeared first.

Negotiators, real-estate agents, lawyers seeking damages, and retailers who show an “original” price before a discount all exploit the same pull. Rational deliberation turns out to be highly sensitive to irrelevant starting points, undercutting economic models that assumed context-free reasoning.

In practice: Subjects who watched a rigged wheel stop at 65 guessed a higher percentage of African UN members than those who saw it stop at 10.

6

Losses outweigh equal gains

People judge outcomes as gains or losses against a reference point, and losses hurt roughly twice as much as equivalent gains please, overturning classical economic theory.

2x how much more intensely losses are felt than equal gains

Prospect theory, published in 1979, was their most consequential result. Traditional economics assumed people maximize expected utility from final wealth. Kahneman and Tversky showed people instead code changes relative to a reference point—usually the status quo—and feel the pain of a loss more intensely than the pleasure of the same-sized gain.

Loss aversion explains why investors cling to losing stocks and refuse fair gambles even when the odds favor them. The Asian disease problem drove the point home: describing a policy as saving 200 of 600 lives made people risk-averse; describing the identical policy as 400 deaths made them risk-seeking.

Preferences are therefore constructed in the moment by framing, not fixed by underlying values. The theory was rigorous enough to reshape economics; Kahneman received a share of the 2002 Nobel Prize in Economics for it.

In practice: In the Asian disease problem the same numbers produced opposite choices depending on whether they were framed as lives saved or lives lost.

7

Expertise does not immunize

Lab-discovered biases carry real financial and human costs in sports, medicine, and forecasting because experts remain vulnerable to the same shortcuts.

DDaryl MoreyRockets GM who replaced look-alike scouting anchors with statistical production DDonald RedelmeierPhysician who demonstrated doctors overweight recent dramatic cases

Lewis follows the research out of journals and into professional settings. Houston Rockets general manager Daryl Morey saw scouts anchoring on how a player looked on tape or resembled past stars instead of on statistical production, systematically overvaluing certain body types.

Physician Donald Redelmeier, who trained under Tversky, showed doctors overweighting a recent dramatic case when judging a new patient’s risk—the same representativeness and availability errors ordinary people make. Military and intelligence forecasters displayed the identical pattern: confident veteran predictions proved no more accurate than simple statistical models.

Expertise can actually increase overconfidence, because experts trust intuitions that once worked in similar-seeming situations. Organizations willing to force judgment against base rates and data improved drafting, diagnosis, and risk evaluation.

In practice: Daryl Morey rebuilt Rockets scouting to counter the tendency to overvalue players who looked like past stars rather than those who produced.

8

Near misses rewrite the past

The mind spends enormous effort constructing alternate versions of events that already happened, shaping regret and blame more than the actual probability of what occurred.

Clear miss Hour-late traveler feels less regret; hard to imagine the outcome undone

Near miss Five-minute-late traveler feels far worse; easy to mentally rewrite

Tversky’s later simulation heuristic, developed partly with Kahneman, describes the tendency to undo the past in imagination. A traveler who misses a flight by five minutes feels far worse than one who misses it by an hour, because the narrow miss is easier to mentally reverse.

The same asymmetry governs blame: an accident caused by a deviation from routine feels more tragic and more culpable than the identical accident caused by routine behavior, simply because the deviation is easier to imagine as avoidable.

Judgment is therefore distorted looking backward as well as forward. The past feels more fluid and reversible than it is, which is why people struggle to accept randomness even after the outcome is fixed. Lewis treats the idea as the emotional core of their work even though it received less experimental development than the earlier heuristics.

In practice: Missing a flight by five minutes produces far more regret than missing it by an hour, even though both travelers end up in the same place.

9

Credit fractures the pair

Professional fame and unequal public credit can dissolve even the most productive scientific partnership and illustrate the very attribution biases the partners studied.

By the early 1980s the collaboration that had produced some of social science’s most cited papers began to fray. Tversky’s individual reputation grew faster; he was the more forceful public presence and collected honors that sometimes overshadowed Kahneman’s share of jointly authored work.

Kahneman, already prone to self-doubt, felt erased from ideas he considered equally his. Colleagues took sides, some treating Tversky as the senior partner and Kahneman as useful support—a perception neither had intended nor fully corrected. The friendship cooled into near silence; each man’s later solo work never matched what they had built together.

Lewis offers no simple villains. Ordinary needs for recognition undid a partnership built on mutual respect. The irony was lost on neither man: two scientists who had mapped how people misattribute causes became subjects of the same distortion.

In practice: Colleagues increasingly treated Tversky as the senior mind and Kahneman as supporting cast, accelerating the personal rupture.

10

The prize comes after death

Recognition often arrives too late for the person who needed it most, complicating any tidy story of scientific legacy.

In 1996 Tversky was diagnosed with metastatic melanoma and given six months. He spent them working, publishing, and quietly repairing pieces of the old friendship, though the earlier warmth never fully returned. He died at fifty-nine.

In 2002 the Nobel committee awarded the Prize in Economic Sciences to Kahneman for the judgment and decision-making research the two had built over three decades. Because the prize is never given posthumously, Tversky’s name could not appear even though the citation rested on work neither could have produced alone.

Kahneman used his speeches and later writing to insist the honor belonged to both and kept describing the findings in the language of their partnership. The gap between joint labor and individual glory became one more case of the unfairness they had spent their careers documenting.

In practice: The 2002 Nobel went to Kahneman alone because Tversky had died in 1996 and the prize cannot be awarded posthumously.

Key terms representativeness heuristic Judging probability by how well something matches a stereotype instead of by actual base rates conjunction fallacy Rating a more specific combination of traits as likelier than one of the traits alone availability heuristic Judging frequency or risk by how easily examples come to mind anchoring effect Letting an arbitrary early number pull later numeric estimates toward it prospect theory Model showing people evaluate outcomes as gains or losses against a reference point rather than final wealth loss aversion The tendency for losses to hurt roughly twice as much as equal gains please illusion of validity Feeling highly confident in a prediction that has almost no actual accuracy simulation heuristic Mentally undoing past events, especially near misses, which intensifies regret and blame framing effects Identical statistics producing opposite choices depending on whether they are described as gains or losses

What to do

Write down your own first estimate before you hear anyone else’s number and notice how far the next figure pulls you. State the relevant base rate out loud before you explain why this particular person or investment feels different. Rewrite any choice framed as something to avoid so it instead names what you would gain, then check whether your preference flips. Agree in advance with a close collaborator on how credit will be split before success makes the question personal. When a vivid recent case dominates a discussion, deliberately list quieter statistical outcomes that rarely make the news.

Questions to think with Where in your own work does confidence travel farther than accuracy, and what simple record could expose the gap? Which recent decision of yours was shaped more by the first number you heard than by the underlying data? If a key partnership in your life ended, how much of the split turned on unequal public credit rather than real contribution? When you feel intense regret, is it mostly about a near miss you can easily imagine undoing? How would your risk choices change if every loss frame were rewritten as a gain frame before you answered?

The other side

Lewis is a narrative journalist, not a psychologist, so the book leans heavily on Kahneman’s retrospective account; Tversky died in 1996 and could not offer a competing version of the friendship’s decline. Readers seeking the actual mathematics of prospect theory’s value function and probability weighting will find the formal model largely skipped. The book stays largely uncritical of the heuristics-and-biases program itself. Gerd Gigerenzer and others have long argued that many classic errors dissolve or reverse when problems use natural frequencies or resemble real-world tasks, suggesting the heuristics may be adaptive rather than defective; Lewis does not engage that debate. The broader replication crisis in psychology also cast doubt on some judgment studies of the era, though loss aversion, framing, and anchoring have held up better than most. Finally, the title concept of undoing and counterfactual simulation receives far less development than the earlier heuristics and functions more as a poetic frame than a fully worked argument.

Who it's for

Readers of behavioral economics who want the human story behind the theories, managers and analysts who rely on expert judgment and need to know its limits, and anyone curious how a close intellectual friendship can produce ideas neither person could reach alone will get the most from the full book.

Read the full summary in Library →