Calibration Certificate No. base-rate-fallacy

Base Rate Fallacy

People ignore general prevalence (base rate) information and rely instead on individuating details or stereotypical information when making probability judgments.

systematic

documented offset from the normative answer

Status — mixedgrounding ± strong

Base-rate neglect has been replicated hundreds of times across different populations and problem formats. However, its robustness is contested: it largely disappears with natural frequency formats (Gigerenzer & Hoffrage, 1995), with causally interpreted base rates, and in between-subjects designs where participants have implicit statistical knowledge. The effect appears to be moderated strongly by information format and task representation.

Calibration record

Lawyer-engineer problemnot reportedParticipants' judgments were dominated by the perceived representativeness of the personality sketch and were largely insensitive to the base-rate proportions. When a sketch was stereotypically 'engineering,' participants gave high probability estimates even when engineers were the minority group. — Kahneman & Tversky, 1973 · n not clearly reported in original paper

Taxi-cab problemmedian estimate = 0.80 vs. Bayesian correct = 0.41The median and modal response was 80%, matching the witness's reliability rather than the Bayesian posterior. Through numerous variations (order of presentation, different colors, different witness credibility), the base-rate neglect remained robust. — Bar-Hillel, 1980 · approximately 350 participants across multiple studies

Source of systematic error

Base-rate neglect is typically explained by the representativeness heuristic: people categorize events based on how representative they are of a prototype, ignoring prior probabilities. Bar-Hillel (1980) proposed a 'relevance' account where individuating information is perceived as more causally relevant to the specific case than abstract statistical base rates. The causal structure of information appears critical - base rates are used more when they can be given a causal interpretation (Tversky & Kahneman, 1980). More recently, some researchers argue that the phenomenon reflects rational Bayesian inference under uncertainty about task structure rather than a cognitive fallacy.

Gigerenzer and Hoffrage (1995) showed that presenting information as natural frequencies (e.g., '8 out of 100') rather than probabilities dramatically reduces or eliminates base-rate neglect, suggesting the fallacy is an artifact of information representation rather than a deep cognitive limitation. Koehler (1996) argued the phenomenon is less universal than claimed and depends heavily on task structure.

Recalibration procedure

Present probabilistic information as natural frequencies (e.g., '8 out of 100') rather than single-event probabilities. Use visual aids such as tree diagrams or contingency tables that make the nested-set structure transparent.

Gigerenzer and Hoffrage (1995) showed that natural frequency formats reduce or eliminate base-rate neglect across multiple problems. Sedlmeier and Gigerenzer (2001) confirmed this with training studies showing that people can learn to reason Bayesianly with frequency representations.

Cross-calibrated against

Representativeness HeuristicparentBase-rate neglect was originally explained as a consequence of the representativeness heuristic, where people judge probability by similarity to a prototype rather than by prior frequency.
Prosecutor's Fallacymechanistically-linkedBoth involve conflating P(evidence|hypothesis) with P(hypothesis|evidence), and both played a role in the Sally Clark case.

Open the same case on your own draft.