A conceptual consideration of the free energy principle in cognitive maps: How cognitive maps help reduce surprise

Abstract There is a nascent view that information processing in the brain can be characterized as Bayesian inference or predictive coding. The mechanics behind this results in the extension that the brain seeks to reduce its entropy and the experience of surprise over its environmental interactions. Critically, the manner in which surprise is reduced involves the generation of Bayesian priors in brain operations, which are ultimately hypotheses about the causes of its sensory and other internal experiences. In this chapter, we seek to annotate these recent theories of the brain as a statistical machine, reviewing the relevant concepts from physics, information science, and cognitive neuroscience. We propose that Bayesian priors specifically take the form of cognitive maps, which are coordinate transformation metrics between neural activity patterns and physical phenomena. We suggest such a view forms the basis for more precise future studies on the nature of information processes in neural networks.