Cookies on this website
We use cookies to ensure that we give you the best experience on our website. If you click 'Continue' we'll assume that you are happy to receive all cookies and you won't see this message again. Click 'Find out more' for information on how to change your cookie settings.

Coding scheme and precision of probabilistic inference as critical source of suboptimality in human decision-making

Making decisions often requires combining multiple pieces of ambiguous or conflicting information from external cues. In such conditions, human choices derive from covert mental operations which resemble Bayesian probabilistic inference, but exhibit a suboptimal variability whose origin remains poorly understood. In particular, it is unclear which fraction of this variability is attributable to the coding scheme and precision of inference and not to sensory processing or response selection, and how much is caused by biases at these different stages of the decision process. We recently developed a computational framework for characterizing the origin and structure of choice suboptimality arising during probabilistic inference. Modeling human choices under this framework in a series of behavioral experiments revealed that a dominant fraction of choice suboptimality reflects the limited coding precision of inference rather than noisy sensory processing or 'probability-matching' response selection. Manipulating the ability to perform inference using stimulus-, choice- or response-based representations further identified a low-dimensional, choice-centered coding scheme for probabilistic inference. These findings set a previously unsuspected bound on the accuracy and consistency of human decisions, but also make selective predictions regarding decision confidence. I will present behavioral and pupillometric data which validate these predictions and inform alterations of decision-making observed in a pharmacological model of early psychosis in schizophrenia.

host: Chris Summerfield