Attention, Discernment, and Bayesian Rationality: Evidential Access and the Recursive Stabilization of Belief

Bayesian epistemology provides a formal account of how rational belief should be revised in response to evidence. Within this framework, beliefs are understood as probabilistic and continuously updated as new information becomes available. However, Bayesian updating presupposes the availability of evidence and does not explain how evidence becomes cognitively accessible in the first place. This paper addresses this foundational question by identifying two complementary cognitive mechanisms that regulate evidential access: attention and discernment. Attention selects environmental signals for cognitive processing, while discernment evaluates their epistemic worth, determining which attended signals constitute genuine evidence warranting belief revision. Bayesian updating governs how beliefs change in response to evidence, but attention and discernment together govern whether and which evidence enters the updating process. Building on a cognitive stabilization framework integrating attention, context, and logic (Matta, 2026a), the paper argues that belief stabilization emerges from the recursive interaction between attentional selection, evaluative discernment, and probabilistic updating. Cognitive stabilization is defined as the progressive convergence of belief toward coherent and environmentally aligned probabilistic representations through iterative updating cycles. Critically, stabilized beliefs in turn refine the discernment and attentional capacities that regulate subsequent evidence selection, producing a recursive and developmental architecture of rational cognition. The paper situates this account in relation to predictive processing, Bayesian brain theories, and alternative models of belief revision. It clarifies that attention and discernment together regulate evidential availability at a level more fundamental than precision weighting alone. This distinction provides a refined account of rational cognition in finite agents and offers new explanatory resources for understanding confirmation bias, cognitive rigidity, and individual differences in learning. The proposed framework integrates epistemology, cognitive science, virtue epistemology, and philosophy of mind, offering a unified account of how belief evolves, stabilizes, and remains responsive to evidence.