Tom W’s Specialty

Kahneman uses the famous 'Tom W' problem to introduce the representativeness heuristic and its companion error, base-rate neglect. Participants read a personality sketch of Tom W — intelligent but unimaginative, orderly, with a need for tidiness and a fondness for science fiction — and were asked to rank his likely field of graduate study. They ranked the fields by how well Tom matched the stereotype, judging him most likely to study computer science or engineering, and ignored entirely how many students are actually enrolled in each field.

The mechanism is representativeness: judging probability by similarity to a stereotype or prototype. The question 'How probable is it that Tom studies computer science?' is hard, so System 1 substitutes the easier 'How well does Tom resemble my image of a computer scientist?' and answers that. The substitution ignores the base rate — the prior probability set by how common each field is — even though the base rate is decisive when the evidence (here, a vague and possibly unreliable sketch) is weak.

Kahneman frames the correct approach in Bayesian terms: a sound judgment anchors on the base rate and adjusts it according to the diagnostic value of the new evidence. If a field has very few students, even strong stereotype-matching evidence should leave its probability modest; if the personality sketch is flimsy or could have been written about many people, it should barely move the base rate at all. Intuition does the opposite — it seizes on the resemblance and discards the prior, producing confident judgments that violate the logic of probability.

The consequence is a pervasive error in prediction and diagnosis. Whenever we judge how likely someone is to belong to a category by how typical they seem, we neglect how common the category is — overpredicting rare types that fit a vivid stereotype and underpredicting common types that do not. The same mistake drives stereotype-based hiring judgments, clinical misdiagnoses, and forecasts that favor the dramatic-but-rare over the mundane-but-frequent.

The applied takeaway is to ask 'how common is this, before I consider the specific evidence?' and to weight the answer according to how reliable that evidence actually is. When a person, product, or situation strikes you as a perfect fit for some category, resist concluding it is therefore probable; check the base rate, and discount vivid but weak evidence. Similarity is seductive, but probability depends on priors that similarity ignores.

Kahneman's deeper observation is that representativeness often produces judgments that feel compelling precisely because they are coherent — Tom W 'just sounds like' a computer scientist — while being statistically wrong. Coherence and probability are different things, and System 1 conflates them. The discipline of base-rate thinking is unnatural and effortful, requiring System 2 to override a vivid, satisfying intuition with a dry statistical correction. That difficulty is exactly why base-rate neglect is one of the most stubborn and consequential biases in human judgment.