CLASSROOM DYNAMICS OF AI-SUPPORTED AUTOMATED WRITING EVALUATION IN AN EFL HYBRID FEEDBACK LOOP

Abstract

This action research study examined how an AI-supported automated writing evaluation (AISAWE) tool can be integrated into EFL writing instruction through a hybrid AI+teacher feedback loop. Because classroom use of AISAWE can introduce feedback overload, interpretation problems, and concerns about learner agency, we implemented an eight-week exploratory action research (EAR) cycle to adapt instructional routines while documenting classroom uptake in a naturalistic setting. Data were primarily qualitative (field notes, teacher focus group data, and student reflections), supplemented by quantitative indicators (pre–post rubric scores, self-report scales, and the System Usability Scale) to contextualize patterns of change. Findings address three research questions. First, uptake was shaped by three recurring practices: managing feedback load, clarifying AI feedback through teacher mediation, and regulating reliance on AI feedback. Second, indicators showed gains in micro-level linguistic accuracy alongside stability in rhetorical structure across the cycle. Third, while students valued speed and coverage, engagement was constrained by usability friction, digital fluency gaps, and motivational and voice-related concerns. Overall, results suggest that AISAWE can support micro-level revision when teacher mediation structures dense feedback and protects learner authorship.