ABSTRACT
Bayesian cognitive science constructs detailed mathematical models of perception, motor control, and many other psychological domains. The models postulate mental activity that approximately conforms to Bayesian norms. Often, the postulated activity is subpersonal. I defend a realist stance towards Bayesian cognitive science. Bayesian models vary widely in their scientific merit, but many are well-confirmed and explanatorily superior to non-Bayesian alternatives. We have good reason to believe that these models are approximately true. I argue that realism about Bayesian cognitive science offers significant explanatory advantages over a rival instrumentalist view, on which Bayesian models are predictive tools that we should not construe even semi-literally.
