Conjunction Fallacy

The conjunction fallacy is a cognitive bias in which people judge a conjunction of two events (A and B) to be more probable than one of the events (A) alone — a violation of the basic laws of probability. The classic demonstration, by Amos Tversky and Daniel Kahneman (1983), uses the Linda problem: subjects are told that Linda is a bank teller who is active in the feminist movement, and then asked whether it is more probable that Linda is (a) a bank teller or (b) a bank teller who is active in the feminist movement. Most subjects choose (b), violating the conjunction rule P(A & B) ≤ P(A).\n\nThe standard interpretation in the heuristics and biases literature treats the fallacy as a failure of statistical reasoning — a departure from normative rationality produced by the representativeness heuristic. Linda's description is more *representative* of a feminist bank teller than of a bank teller simpliciter, and subjects confuse representativeness with probability.\n\nBut from the perspective of adaptive cognition, the conjunction fallacy is not a broken mechanism operating in a vacuum. It is a well-calibrated heuristic operating in the wrong ecology. In natural social environments, detailed descriptions often carry genuine diagnostic value: someone who matches a detailed profile is often more *likely* to belong to a specific social category than to a general one, because the detailed description conveys membership in a community where the traits co-occur. The 'fallacy' is a mechanism evolved for social inference being deployed on a stripped-down probability puzzle where the details are explicitly declared irrelevant.\n\n\n\n\nThe conjunction fallacy is not a bug in human reasoning. It is a social inference engine being tested on a problem that only looks like probability because someone removed all the context that makes social reasoning work.