Maurice Allais challenged the world's leading decision theorists — at a conference in 1952 — with a simple choice problem. Most of them, including Leonard Savage (the author of expected utility theory), violated their own theory. The embarrassment launched a revolution.

The Framework

Allais's paradox presents two pairs of gambles where most people's preferences violate the independence axiom of expected utility theory. In the first pair, people prefer a sure $500K over a gamble with an 89% chance of $500K, 10% chance of $2.5M, and 1% chance of nothing. In the second pair (with modified probabilities), people's preferences reverse in a way that's mathematically inconsistent with expected utility.

The violation is driven by the certainty effect: in the first pair, the sure thing offers the qualitative leap from 'probable' to 'certain,' which carries a premium far exceeding its probability-weighted value. In the second pair, both options are gambles (neither is certain), so the certainty premium doesn't apply, and preferences follow expected value. The inconsistency proves that human preferences don't satisfy expected utility's axioms — a founding evidence for prospect theory.

Where It Comes From

Chapter 29 of Thinking, Fast and Slow presents Allais's paradox as the first major challenge to expected utility theory. Allais deliberately embarrassed Savage at the 1952 conference by showing that Savage's own choices violated his own theory. The anecdote demonstrates theory-induced blindness: Savage initially couldn't see the violation in his own behavior.

> "Allais set a clever trap for expected utility theorists at a conference in 1952." — Thinking, Fast and Slow, Ch 29

The Implementation Playbook

Offer Design: The certainty effect means 'guaranteed' commands a premium over 'highly probable' that exceeds the mathematical difference. '100% money-back guarantee' is worth far more than '95% satisfaction rate' — not because of the 5% difference, but because certainty is qualitatively different from near-certainty.

Risk Assessment: Your stakeholders' preferences will be inconsistent across different choice framings. A board that chooses the 'safe' option in one context may choose the 'risky' option in a mathematically equivalent reformulation. Present both framings and discuss the inconsistency.

Key Takeaway

Allais's paradox proved that expected utility theory's axioms don't describe human behavior — they prescribe an ideal that humans systematically violate. The certainty effect, formalized 30 years later in prospect theory, explains the violation: the qualitative leap from probable to certain carries psychological value that no probability-weighted calculation captures.

Continue Exploring

[[Certainty Effect]] — The psychological mechanism behind Allais's paradox

[[Expected Utility Theory (Bernoulli)]] — The theory that Allais's paradox challenges

[[Theory-Induced Blindness]] — Why Savage couldn't see the violation in his own choices

📚 From Thinking, Fast and Slow by Daniel Kahneman — Get the book