Daniel Kahneman, answered from the texts and cited to the page.
The representativeness heuristic is what happens when the mind answers a hard question by substituting an easier one. When you ask people how probable it is that Linda is a bank teller, they answer instead — without knowing they have switched — how much Linda resembles the prototype of a bank teller. The substitution is automatic, and its consequences are systematic.1
The Linda problem was designed precisely to expose this. Amos and I constructed her description to be representative of an active feminist and unrepresentative of a bank teller.2 The critical test was whether people would rank "Linda is a bank teller and active in the feminist movement" as more probable than "Linda is a bank teller." By the logic of probability, the conjunction cannot exceed its constituent — every feminist bank teller is also a bank teller, so the set of feminist bank tellers is wholly contained within the set of bank tellers.3
And yet 89 percent of respondents in the probability condition ranked the conjunction higher.4 What makes the finding compelling is the parallel with the similarity condition. When a separate group ranked the same outcomes by resemblance rather than probability, 85 percent showed the same ordering.5 The correlation between mean probability rankings and mean similarity rankings was.99.6
That near-identity is the signature of attribute substitution: the two instructions produced virtually the same responses because respondents were, in effect, answering the same question — the accessible one about resemblance — regardless of which question they had been asked. The heuristic is not a failure of intelligence. The similarity judgment is reasonable.
Linda does resemble a feminist bank teller more than she resembles the stereotype of bank tellers.7 What goes wrong is the transfer of that similarity ordering into the domain of probability, where it violates a basic logical constraint. People do not defend the ranking when the violation is pointed out to them; they withdraw it. That is the clearest evidence that they endorse the normative standard — they simply did not apply it, because representativeness got there first.
The respondents are therefore expected to substitute a judgment of similarity (representativeness) for the required judgment of probability.Maps of Bounded Rationality, Psychology for Behavioral Economics, pp. 15–16
The description of Linda was constructed to be representative of an active feminist (F) and unrepresentative of a bank teller (T).Judgment under Uncertainty, the Kahneman and Tversky Chapters, pp. 77–78
The set of feminist bank tellers is wholly included in the set of bank tellers, as every feminist bank teller is also a bank teller. Therefore the probability that Linda is a feminist bank teller must be lower than the probability of her being a bank teller.Thinking, Fast and Slow
it is much more problematic that 89 percent of respondents in the probability group also ranked the conjunction higher than its constituent.Maps of Bounded Rationality, Psychology for Behavioral Economics, pp. 15–16
85 percent of respondents in the similarity group ranked the conjunction item (#8) higher than its constituent, indicating that Linda resembles the image of a feminist bank teller more than she resembles a stereotypical bank teller.Maps of Bounded Rationality, Psychology for Behavioral Economics, pp. 15–16
The correlation between the mean rankings was.99. Furthermore, the proportion of respondents who ranked the conjunction (item #8) higher than its constituent (#6) was about the same for representativeness (85%) and for probability (89%).Maps of Bounded Rationality, Nobel Prize Lecture, p. 20
Everyone agrees that Linda fits the idea of a 'feminist bank teller' better than she fits the stereotype of bank tellers. The stereotypical bank teller is not a feminist activist, and adding that detail to the description makes for a more coherent story.Thinking, Fast and Slow