Judgment Under Uncertainty: Heuristics and Biases

The book introduced and elaborated three heuristics: representativeness, availability, and anchoring. Representativeness governs probability judgments about category membership and causal sequences; availability governs frequency and likelihood estimates based on ease of recall; anchoring describes insufficient adjustment from initial values. The volume's argument replaced the assumption that deviations from probability theory reflect random noise or motivational bias. Instead, it proposed that these heuristics are cognitively efficient strategies that work well in many contexts but produce systematic errors in specific experimental settings. The 1974 Science paper by Tversky and Kahneman provided the initial statement; the 1982 book extended it across domains including medical diagnosis, legal judgment, and social attribution.

Thinking, Fast and Slow (2011) is a synthesis and popularization of the research program this 1982 volume consolidated. Kahneman's later book organizes the heuristics and biases findings under the System 1/System 2 framework, a conceptual architecture absent from the 1982 collection. The edited volume presents individual papers on representativeness, availability, and overconfidence without the unifying dual-process narrative that Kahneman developed later. Reading Judgment Under Uncertainty reveals the empirical foundations and technical arguments in their original form, including papers by contributors beyond Kahneman and Tversky, such as Bar-Hillel on representativeness, Ross on attribution, and Lichtenstein on calibration.

Gerd Gigerenzer and colleagues at the ABC Research Group offered the most sustained critique, arguing in works like Simple Heuristics That Make Us Smart (1999) that the heuristics and biases framework evaluates judgment against content-blind normative standards that ignore environmental structure. Gigerenzer contended that heuristics should be assessed by ecological rationality—whether they perform well in real-world environments—rather than by coherence with probability axioms. He also demonstrated that reformulating probability problems in frequency terms dramatically improves performance, suggesting that apparent biases reflect experimental format rather than inherent cognitive limitations. This debate reshaped how researchers frame the relationship between heuristics and rationality.

The editors designed the book to function as a comprehensive statement of the program rather than an archival anthology. About half the chapters are edited versions of previously published articles; the remainder were written specifically for the volume. These new chapters extended the heuristics and biases approach into areas not covered by the original papers: medical diagnosis, clinical reasoning, risk perception, debiasing techniques, and the psychology of hindsight. The new material also addressed methodological issues and attempted to respond to early criticisms. The result is a volume that functions as both a primary-source collection and an original contribution, with chapters organized into sections on representativeness, causality, availability, covariation, overconfidence, and corrective procedures.

The conjunction fallacy describes the finding that people judge a conjunction of two events as more probable than one of its constituent events, violating a basic axiom of probability theory. The classic demonstration involves a description of a person that makes a specific combination of traits seem more representative than a broader category. In this volume, the fallacy appears in the representativeness section as evidence that probability judgments are driven by resemblance to prototypes rather than by set-theoretic logic. The finding matters because it cannot be explained by random error or motivational bias; it reflects a systematic property of how representativeness operates. The conjunction fallacy later became one of the most cited phenomena in the heuristics and biases literature.

The overconfidence section of the volume includes chapters on calibration of probabilities and case-study judgments, documenting that people tend to be more confident than accurate. Subsequent research has refined and partially challenged these findings. Some studies suggest that overconfidence is domain-dependent, with experts in some fields showing better calibration than general populations. Others have argued that the standard experimental paradigms elicit overconfidence through task design rather than reflecting a general cognitive tendency. The 'hard-easy effect'—where overconfidence increases on difficult tasks and decreases on easy ones—has been extensively documented. The volume's chapters by Lichtenstein, Fischhoff, and Phillips, and by Oskamp, remain the foundational statements that later work has extended and qualified.