Phase Gated Bayesian Cognition

This article introduces the framework of Phase Gated Bayesian Cognition (PGBC), a computational model that addresses how the brain manages high-entropy sensory input without overwhelming its internal priors. While traditional Bayesian Brain models often assume a continuous flow of evidence updating, PGBC proposes a "gating" mechanism that segments cognitive processing into distinct phases. By utilizing a phase-gated architecture, the model explains how the cognitive system can switch between high-plasticity phases (where sensory evidence heavily updates internal models) and stability phases (where priors are protected from noise). This approach provides a novel perspective on the "stability-plasticity" dilemma and offers potential insights into the neurocomputational foundations of attention, belief revision, and certain psychiatric conditions where gating mechanisms may be impaired. The paper explores the mathematical formalization of these gates and discusses their alignment with known neurobiological rhythms, specifically how oscillatory phases may serve as the physical substrate for these Bayesian gates.