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

AI systems increasingly mediate how individuals form beliefs and
construct knowledge, yet current fairness and accountability frame-
works inadequately address the cognitive and epistemic consequences
of prolonged AI interaction. Building on theories of epistemic in-
justice [46, 92], metacognition [42], and identity formation [37], we
argue 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-minimizing
interfaces undermine reflective deliberation [66]. We propose a
metacognitive restoration framework with three design principles:
deliberative scaffolding, cognitive resistance preservation, and epis-
temic authorship support. Our work contributes (1) a conceptual
framework linking AI design to cognitive autonomy, (2) empiri-
cally grounded design recommendations, and (3) evaluation met-
rics prioritizing epistemic resilience. This research advances FAccT
scholarship by centering long-term cognitive impacts as an account-
ability concern alongside traditional fairness metrics [14, 113].