Allais Paradox

The Allais Paradox is the foundational challenge to subjective expected utility theory, discovered by the French economist Maurice Allais in 1953. Allais presented a choice problem in which most people violate the independence axiom that underlies Leonard Jimmie Savage's derivation of expected utility. The paradox shows that when choices involve certainty, people prefer guaranteed outcomes over probabilistically equivalent gambles — a pattern that the Savage framework labels irrational but that descriptive decision theory has come to treat as a stable feature of human preference. The paradox is not a trick question. It is a demonstration that the axioms of rational choice embed a particular metaphysics of uncertainty — one that treats probabilistic mixtures as interchangeable regardless of their component parts. The Allais violations suggest that people value certainty as a distinct property, not merely as a probability of one. This insight has shaped behavioral economics, prospect theory, and the current debate over whether AI systems trained on Savage-style objectives inherit the same blind spots about human risk preference that the axioms themselves encode. The paradox is not resolved. It is suppressed — by a field that prefers elegant formalism to descriptive honesty.