The Bayesian Savant

Computational psychiatry promises a fresh and formal approach to mental health, and autism has become its so-called poster child. Key concepts from computational neuroscience are now finding their way into discussions about the pathophysiology and psychopathology of autism spectrum disorder (ASD) (1, 2, 3, 4). This is exemplified beautifully by Sevgi et al. (5), who report that “higher autistic traits in healthy subjects are related to lower scores in a learning task that requires social cue integration.” Careful Bayesian modeling of this learning suggests that trait-related differences are not explained by a failure to process social stimuli per se, but rather by the extent to which participants afford precision to—or attend—social cues. So why is it important? For people unfamiliar with things like the Bayesian brain and precision, we start with a brief review of the ideas that motivated Sevgi et al. (5)