DR@W

Standard models of decision-making capture regularities in risk-taking and delay-discounting by applying subjective transformations to objective choice primitives.

Such subjective transformations are typically thought of as capturing stable ‘preferences’. In a stark departure from the standard approach, noisy cognition models represent behavior as the outgrowth of optimal reactions to noisy perceptions and re-combinations of choice primitives. Here, we test the predictions of the two model classes against each other by systematically varying presentation formats of identical choice primitives in ways we expect to affect cognitive noise. The results illustrate that behavioral regularities such as insensitivity to probabilities and to time delays can be systematically shifted and even reversed by subtle alterations in the presentation of identical choice tasks. Canonical patterns such as risk aversion increasing in stakes and delay-discounting declining in stakes obtain when transparently changing the numerical units in which rewards are expressed while keeping the underlying stakes constant. These results appear puzzling from the perspective of standard models. They do, however, precisely track the predictions emerging from noisy cognition models. The results illustrate the generative and hence causal meaning of noisy cognition parameters, which provide a stylized representation of information processing by the brain. The implication is that treating choices as dissociated from physical processes in the brain, as advocated in standard economics, risks overlooking the key to understanding — and ultimately predicting — behavior.