Bayesian Theories, Visual Perceptual States, and Multiple Realizability

Bayesian accounts of perception propose that the brain combines prior expectations with sensory data to generate hypotheses that best approximate the world. Within the last two decades, they have been gaining considerable traction in cognitive neuroscience, providing compelling explanations of many aspects of perceptual cognition (see Griffiths et al., 2010). However, they also have important implications for key debates in the Philosophy of Mind. Namely, this article proposes that they potentially offer a way to explain perceptual states without making any specific commitments regarding the biological hardware that produces them, thereby ensuring multiple realisability and avoiding chauvinism-an unintuitive implication faced by other theories, that only humans or other vertebrate creatures have perceptual states. This is because Bayesian accounts of perception are defined at Mares (1982) "computational level" of explanation, and a system defined at this level imparts few constraints regarding how its operations are (or could be) implemented. As a result, Bayesianism potentially provides a set of principles that might describe something akin to the universal structure of perceptual states, which could apply even to intelligent AI or extraterrestrial life. This article outlines this point by focusing on visual states as a central example.