What it is
This work analyzes when the components of Bayesian models of perception, namely the prior belief, likelihood function, and loss function, can be recovered from behavioral data. The authors determined the situations under which some components are systematically confounded and identified the experimental-design choices that resolve such ambiguity. Their analytical results guarantee in-principle identifiability under broadly applicable conditions, without any prior knowledge of the prior or the encoding, and simulations and applications on behavioral datasets confirm the theory holds in realistic settings.
Why it matters
Bayesian decision theory is a major normative framework for modeling perception and cognition, yet it has been unclear whether its core components can actually be disentangled from behavior, which undercuts claims built on fitted models. Establishing when the prior, likelihood, and loss function are identifiable, and when they are confounded, puts such modeling on firmer ground and shows that reliable recovery often requires data collected at multiple noise levels.
Every metric behind this entry is listed, with its source, under Sources and data below.
Filed underNeural and Behavioral Psychology Studies, Neural dynamics and brain function, Visual perception and processing mechanisms