The Heuristics and Biases Audit
Systematically identify and correct the mental shortcuts that distort judgment
Kahneman and Tversky identified that when people face difficult judgment questions, System 1 often substitutes an easier related question and answers that instead, without the person being aware of the substitution. When asked about the probability that Steve is a librarian versus a farmer, people rely on how much Steve resembles the stereotype of a librarian (the representativeness heuristic) rather than considering that there are far more farmers than librarians (the base rate). When asked about the frequency of divorces among professors, people search for examples they can recall (the availability heuristic) rather than consulting statistics.
These heuristics are not errors in themselves; they are efficient shortcuts that are usually approximately correct. But they produce systematic, predictable biases in specific circumstances. The representativeness heuristic leads to base-rate neglect, the conjunction fallacy (judging that Linda is more likely to be a feminist bank teller than a bank teller), and insensitivity to sample size. The availability heuristic leads to overestimation of dramatic risks (plane crashes, terrorism) and underestimation of silent killers (diabetes, asthma).
The practical framework involves building systematic audit procedures that check for known biases at predictable decision points. Organizations benefit more from this approach than individuals because they can institutionalize checks and create a vocabulary for identifying errors in real time.
- When faced with a hard question, System 1 substitutes an easier question and answers that instead
- The representativeness heuristic judges probability by similarity to stereotypes, ignoring base rates and sample sizes
- The availability heuristic judges frequency by ease of recall, overweighting vivid, recent, and emotional events
- Regression to the mean is a statistical inevitability that people systematically fail to anticipate, instead constructing causal stories
- Biases are predictable and systematic, meaning they can be anticipated and checked even if they cannot be eliminated
- Name the heuristic in playBuild a vocabulary for common heuristics. When you catch yourself or others making a judgment, ask: are we answering the question that was asked, or a different, easier question? Is this an availability judgment (based on what comes to mind) or a representativeness judgment (based on resemblance to a stereotype)?
- Check for base-rate neglectFor any probability assessment, start by asking: what is the base rate? Before deciding Steve is more likely a librarian than a farmer, check the ratio of librarians to farmers. Before judging a startup's prospects, check the base rate of startup success in that industry.
- Apply the sample-size correctionSmall samples produce extreme results by chance alone. Counties with the lowest cancer rates are small, rural counties, but so are counties with the highest cancer rates. Before drawing conclusions from data, ask whether the sample size is large enough to produce stable results.
- Expect regression to the meanExtreme performance in any direction tends to be followed by performance closer to the average. A spectacular quarter does not predict continued outperformance. An athlete's best season is likely followed by a more ordinary one. Build this expectation into forecasts rather than constructing causal stories about why the trend will continue or reverse.
- Institute organizational checklistsCreate decision checklists that force explicit consideration of base rates, sample sizes, regression to the mean, and alternative explanations. Kahneman argues that organizations are better positioned than individuals to implement these corrections because they can mandate procedures and create cultural norms around quality control.
Kahneman and Tversky described Linda as a 31-year-old former philosophy major, deeply concerned with social justice and discrimination. Participants were asked which was more probable: Linda is a bank teller, or Linda is a bank teller who is active in the feminist movement. The overwhelming majority chose the latter.
The heuristics and biases research program was launched by Kahneman and Tversky in their landmark 1974 Science paper, 'Judgment Under Uncertainty: Heuristics and Biases.' Their insight was that cognitive shortcuts are not random failures of logic but systematic patterns that follow identifiable rules, making errors predictable and, in principle, correctable. The program generated dozens of replicated experiments over four decades and fundamentally changed how psychologists, economists, and policy makers understand human judgment.