Answers are grounded in this paper and cite their sources.
Examines epistemic regulation in junior high science learning, suggesting algorithmic ambiguity enhances knowledge construction.
<ns3:p> Abstract* Background The rapid diffusion of generative artificial intelligence has introduced new epistemic challenges in education, particularly through algorithmically generated content that appears credible yet may contain subtle distortions. This study addresses these challenges by developing a parallel mediation model of epistemic regulation under algorithmic ambiguity, conceptualized as Deepfake Learning Credibility Ambiguity (DLCA). Drawing on the Stimulus–Organism–Response (SOR) framework, DLCA is positioned as a contextual stimulus that activates three regulatory mechanisms—Epistemic Vigilance, AI Verification Competence, and Authenticity Commitment—which function as parallel mediators shaping Authentic Knowledge Construction in junior high school science learning. Methods This study employed a quantitative design using Partial Least Squares Structural Equation Modeling (PLS-SEM) to test the proposed parallel mediation model. Data were collected from 1,237 junior high school students. The structural relationships among DLCA, the three epistemic regulatory mechanisms, and Authentic Knowledge Construction were examined to assess both direct and indirect effects. Results The findings indicate that DLCA significantly predicts Epistemic Vigilance (β = 0.540), AI Verification Competence (β = 0.470), and Authenticity Commitment (β = 0.410). These three mechanisms significantly enhance Authentic Knowledge Construction, with the overall model explaining 63% of its variance (R <ns3:sup>2</ns3:sup> = 0.63). The indirect effects of DLCA on Authentic Knowledge Construction through Epistemic Vigilance (β = 0.157), AI Verification Competence (β = 0.113), and Authenticity Commitment (β = 0.131) are significant, supporting a multidimensional parallel mediation structure. Although a direct effect of DLCA on Authentic Knowledge Construction remains significant (β = 0.120), the primary influence operates through epistemic regulatory pathways. Conclusions The findings suggest that algorithmic ambiguity does not merely create epistemic risk but functions as a catalyst for epistemic adaptation. By activating multidimensional regulatory mechanisms, DLCA fosters authentic knowledge construction. This study contributes to AI-in-education research by offering an integrated explanatory framework for understanding epistemic literacy and promoting responsible AI use in educational contexts. </ns3:p>
No takes yet. Share an insight, caveat, or question.
Prihatni et al. (2026) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: