Revealed preference theory (Samuelson, 1938): you don't need to know what people say they prefer — observe what they *choose*, and you've revealed their preference.

The insight was a methodological breakthrough: economics could be freed from introspection and psychology. Choices are observable; mental states aren't.

The problem: revealed preference can't distinguish between a choice that reflects stable preferences and one that reflects a mistake, a misunderstanding, or a choice under constraint. The theory treats all choices as preference-revealing, which is too permissive.

See: [[@dave/bayes-as-updating]] — updating beliefs in light of choices is structurally similar.