Artificial Intelligence-Assisted Learning and Epistemic Agency: A Critical Scoping Review and Conceptual Synthesis of Psychological Responses and Epistemic Dependence

Artificial intelligence-assisted learning can improve academic performance while simultaneously weakening the learner’s role in understanding, evaluation, and judgment. This study presents a critical scoping review and conceptual synthesis of psychological responses to artificial intelligence-assisted learning and their implications for human epistemic agency. Structured searches of Scopus and the Education Resources Information Center identified ۳۲ empirical studies published between ۲۰۲۲ and ۲۰۲۶. Because ۳۸ of ۱۰۶ reports sought for retrieval were inaccessible, the corpus is treated as a bounded empirical baseline rather than an exhaustive census of the literature. Two broad trajectories emerged. Active and regulated use was associated with higher self-efficacy, confidence, perceived competence, feedback engagement, and metacognitive monitoring. In contrast, unregulated automation was associated with dependency, overreliance, anxiety, self-doubt, and uncertainty about competence. Philosophically, these patterns support a distinction between epistemic scaffolding and epistemic substitution. Drawing on Sosa’s virtue epistemology, scaffolding occurs when artificial intelligence contributes to task accuracy while the learner’s adroitness remains causally relevant to success, preserving apt performance. Substitution occurs when an output is accurate but its success is no longer attributable to the learner’s competence, rendering the performance epistemically inapt. Cognitive offloading helps explain the transition between these modes. No included study directly measured the impostor phenomenon using a validated impostor instrument; it is therefore reframed as a Competence Attribution Ambiguity Hypothesis rather than an established empirical effect. Responsible artificial intelligence use in education requires calibrated dependence, active metacognitive control, and preservation of human epistemic agency.

کلیدواژه ها:

Artificial intelligence ، AI-assisted learning ، Epistemic Agency ، Epistemic dependence ، Psychological responses