Thinking, Fast and Slow

Daniel Kahneman · 2011

Thesis

Human judgment runs mostly on a fast, automatic system that produces confident answers from whatever information is at hand, and its errors are systematic enough to predict and correct.

The main ideas

Two modes of thought: System 1 is fast, automatic and effortless; System 2 is slow, deliberate and lazy, and mostly endorses what System 1 suggests.
System 1 builds the most coherent story it can from the information in view and ignores what is missing. Confidence follows the coherence of the story, not the quality of the evidence.
Faced with a hard question, System 1 answers an easier one and does not notice the swap.
Arbitrary numbers anchor estimates, chance explains more of performance than people credit, and experts are often most confident where they are least valid.
Simple formulas beat expert judgment in most noisy domains. Forecasts improve when they start from the outcomes of similar cases.
People evaluate outcomes as gains and losses from a reference point, and losses weigh about twice as much as gains.
Memory judges an experience by its peak and its end, not its duration, so the self that chooses is not the self that lives through the result.

System 1, System 2 and substitution

Kahneman argues that most judgments come from System 1, which answers instantly, and that System 2 checks those answers only when it is forced to, because checking takes effort. The evidence is the bat-and-ball problem: a bat and ball cost $1.10, and the bat costs $1 more than the ball. The answer that comes to mind is 10 cents; the correct answer is 5 cents. More than 50% of students at Harvard, MIT and Princeton gave the intuitive answer, and more than 80% at less selective universities.

The main way System 1 handles hard questions is substitution: it answers an easier, related question. Asked how happy they are with their life, people report their current mood. In a study of German students, asking "How happy are you?" first and then "How many dates did you have last month?" produced a correlation near zero. Reversing the order made the correlation about as high as correlations between psychological measures get, because the dating question had become the answer to the happiness one.

Implication: when an answer arrives instantly to a hard question, identify which easier question it actually answers before acting on it.

What you see is all there is

Kahneman's name for System 1's central habit is WYSIATI: it builds the best story it can from the information available and treats that information as complete. It does not register missing evidence, so a story built from two facts can feel as solid as one built from twenty, and often more solid, because fewer facts leave fewer contradictions. Subjective confidence measures the coherence of the story, not how much of the relevant evidence has been seen.

WYSIATI explains several of the book's other effects. The halo effect, in which one known trait colors the judgment of all others, works because early information sets the story that later information is fitted into. Overconfidence follows directly: people with little information have little to contradict their conclusions.

Implication: before a decision, list the information you would want and do not have. A conclusion that feels certain on thin evidence is a sign of WYSIATI, not of a strong case.

Anchoring

Any number considered before an estimate pulls the estimate toward it, even when the number is known to be irrelevant. Kahneman describes two mechanisms. System 2 adjusts away from the anchor and stops too early. System 1 treats the anchor as a suggestion and selectively recalls evidence consistent with it.

The evidence: Kahneman and Tversky rigged a wheel of fortune to stop at 10 or 65, then asked people what percentage of African nations were UN members. Average estimates were 25% and 45%. German judges with an average of more than 15 years on the bench rolled dice loaded to show 3 or 9, then sentenced a shoplifter in a hypothetical case: those who rolled 9 gave 8 months on average, those who rolled 3 gave 5.

Implication: in a negotiation, the first number sets the range, so make the first offer or explicitly reject theirs as a reference. Treat any number seen before an estimate, including an asking price, as contamination.

Regression to the mean

Extreme performances tend to be followed by less extreme ones, because any outcome combines skill and luck, and luck does not persist. People see regression and invent a causal explanation for it. Kahneman's case: flight instructors in the Israeli Air Force told him that praising cadets for a smooth landing was followed by a worse one, and criticizing a rough landing was followed by a better one, so punishment worked and praise did not. Kahneman showed that the pattern follows from chance alone. A cadet who has an unusually good landing will probably do worse next time whatever the instructor says. Because praise follows good performance and criticism follows bad, chance alone makes criticism look effective and praise look harmful.

Intuitive forecasts ignore regression: they are as extreme as the evidence. Kahneman's correction is to start from the average outcome and move toward the intuitive prediction only in proportion to how well the evidence correlates with the result.

Implication: after an unusually good or bad quarter, hire or intervention, expect a move back toward the average and do not credit or blame the most recent action for it.

The illusion of validity and the case for formulas

Confidence in a judgment is no evidence of its accuracy. In the Israeli army Kahneman rated officer candidates on a leaderless group exercise; the team felt sure of its assessments, and follow-up showed their predictive power was negligible. Their confidence did not drop when they learned this. In an analysis of a wealth-management firm, Kahneman computed the correlation between each of 25 advisers' rankings in each pair of years over eight years. The average of the 28 correlations was .01: rankings were luck, yet the firm paid bonuses on them.

Kahneman draws on Paul Meehl, whose 1954 review of 20 studies found that simple statistical rules predicted outcomes better than trained clinicians. Of roughly 200 later comparisons, about 60% favor the algorithms and the rest are ties. Formulas win because experts try to be clever, weigh cues inconsistently, and give different answers to the same case on different days. When Kahneman redesigned army interviews, he had interviewers score six traits from 1 to 5 in fixed order, then close their eyes and give an overall intuitive score. That final intuition, disciplined by the structured scores, predicted as well as the sum of the six.

Implication: for repeated judgments such as hiring, credit or triage, write down the few factors that matter, score them separately, and combine them mechanically before letting intuition in.

The planning fallacy and the outside view

Plans and forecasts are built from the specifics of the case in hand, the inside view, and so assume a best case. The correction is the outside view: find the reference class of similar projects and start from its distribution of outcomes. Kahneman's case is his own. His team writing a high-school textbook on judgment and decision making estimated about two years to finish. He asked Seymour Fox, the team's expert on such projects, how long similar teams had taken. Fox said about 40% never finished, and he could not think of one that finished in less than seven years or took more than ten. The team continued anyway. The book took eight years and was never used.

The base rate was known to the team's own expert and changed nothing, so Kahneman concludes the outside view has to be imposed as a procedure. He endorses Gary Klein's premortem: before committing, the team imagines that the project has failed a year later and writes a brief history of the failure.

Implication: for any estimate of cost or time, first ask what happened to similar projects and anchor on that number, then adjust for what is specific to this case.

Prospect theory and loss aversion

Kahneman and Tversky's prospect theory holds that people evaluate outcomes as gains or losses relative to a reference point, not as final states of wealth, and that losses loom larger than gains. The evidence is a coin flip: lose $100 on tails, win X on heads. For many people X must be about $200 before they accept. The loss aversion ratio is usually between 1.5 and 2.5. Sensitivity also diminishes: the gap between $100 and $200 feels larger than the gap between $900 and $1,000.

Combined with the overweighting of small probabilities, this produces the fourfold pattern. People are risk averse for likely gains (they accept a sure settlement below a strong case's value) and risk seeking for likely losses (they gamble to avoid a sure loss). For unlikely outcomes the pattern flips: they buy lottery tickets and overpay for insurance. The second cell is the costly one: a losing position invites desperate bets because accepting the sure loss is painful.

Implication: set the reference point deliberately. Judge a portfolio or a business line by whether you would enter it today, not by gain or loss from the purchase price, and treat a string of small bets as one policy so each loss is not felt alone.

The two selves

Kahneman separates the experiencing self, which lives through each moment, from the remembering self, which keeps score and makes decisions. The remembering self follows the peak-end rule, rating an episode by the average of its worst moment and its end, and it neglects duration. In colonoscopy studies, a patient whose procedure lasted 24 minutes and ended at a low pain level remembered it as less bad than a patient whose 8-minute procedure ended at peak pain, although the first endured more total pain. In the cold-hand experiment, participants held one hand in 14°C water for 60 seconds, and the other hand for 90 seconds, the last 30 with the water warmed by about 1°C. Given a choice of which trial to repeat, about 80% of those who noticed the pain diminish chose the longer one: more pain, better memory.

Because decisions are made on memories, people choose for the remembering self and impose the costs on the experiencing self. Implication: when designing any experience you control, such as a negotiation, a project or a customer process, the ending carries disproportionate weight.

Limits

Several studies in the book have failed to replicate. Kahneman's 2012 email to priming researchers warned of "a train wreck looming," and in 2017 he wrote that he had placed too much faith in underpowered studies in the priming chapter. A 23-lab replication of ego depletion (Hagger et al., 2016, 2,141 participants), which the book presents as established, found an effect near zero.
The Israeli parole study that the book cites, in which favorable rulings fell from about 65% after meals to near zero before them, was challenged by Weinshall-Margel and Shapard (2011), who showed that case order was not random: panels tried to finish a prison's cases before a break, and unrepresented prisoners tended to come last.
System 1 and System 2 are a metaphor, not brain structures, and the book's list of biases is not a unified model. Gigerenzer and others argue that the same heuristics are accurate in the environments they evolved for, and that many errors shrink when problems are stated as frequencies instead of probabilities.

Strangers to Ourselves + Redirect — the unconscious processes behind the fast judgments Kahneman describes, and how little introspection reaches them.
The Laws of Human Nature — the same biases told through historical cases rather than experiments.
Fooled by Randomness — applies the same failures of probabilistic judgment to trading and track records.