<![CDATA[In English as a Foreign Language (EFL) writing instruction, the pedagogical value of generative AI depends less on technological capability than on how learners are guided to evaluate, regulate, and respond to AI-mediated feedback. Framed within assessment as learning, this study examines how Internet Epistemic Beliefs (IEB) shape students’ Writing Perceptions (WP) directly and indirectly through Self-Assessment (SA) and perceptions of ChatGPT as a writing self-assessment tool (GPT). A cross-sectional survey was conducted among 223 Indonesian EFL university students with prior experience using ChatGPT for academic writing. Composite scores from validated measures of IEB, SA, GPT, and WP were analyzed using theory-driven path analysis with regression procedures. The results indicate that the model significantly explained variance in WP (F = 60.46, p < .001). SA emerged as the strongest predictor of WP (β = 0.43, p < .001), underscoring the central role of reflective judgment in writing development. IEB also exerted a significant direct effect on WP (β = 0.28, p < .001) and strongly predicted SA (β = 0.53, p < .001). Although IEB significantly predicted perceptions of ChatGPT (β = 0.44, p < .001), GPT did not independently predict WP when epistemic beliefs and self-assessment were controlled (β = 0.06, p = .27). Mediation analysis confirmed that IEB influenced WP primarily through SA rather than through AI feedback perceptions. Pedagogically, the findings highlight the importance of prioritizing epistemic awareness and structured self-assessment when integrating AI tools into EFL writing classrooms.]]>
Copyrights © 2026
