Algorithmic Epistemic Environments and Metacognitive Agency: A Framework for Evaluating Cognitive Impacts of AI Systems

AI systems increasingly mediate how individuals form beliefs andconstruct knowledge, yet current fairness and accountability frame-works inadequately address the cognitive and epistemic consequencesof prolonged AI interaction. Building on theories of epistemic in-justice [46, 92], metacognition [42], and identity formation [37], weargue that AI-mediated information environments can systemati-cally erode users’ capacity for autonomous belief formation—a con-dition we term epistemic displacement. Through qualitative anal-ysis of design patterns in recommender systems, search engines,and conversational AI, we demonstrate how friction-minimizinginterfaces undermine reflective deliberation [66]. We propose ametacognitive restoration framework with three design principles:deliberative scaffolding, cognitive resistance preservation, and epis-temic authorship support. Our work contributes (1) a conceptualframework linking AI design to cognitive autonomy, (2) empiri-cally grounded design recommendations, and (3) evaluation met-rics prioritizing epistemic resilience. This research advances FAccTscholarship by centering long-term cognitive impacts as an account-ability concern alongside traditional fairness metrics [14, 113].