- Oct 20, 2026
- Posted by:
- Category: Abstract of 10th-educationconf
Abstract Book of the 10th International Conference on Advanced Research in Education
Year: 2026
[PDF]
AI-Mediated Corrective Feedback and EFL Writing Accuracy: A Study of Algerian University Learners
Mohamed Sallaye
ABSTRACT:
This study investigated the effect of AI-mediated corrective feedback, delivered through ChatGPT, on the writing accuracy of Algerian university EFL learners, and explored how learners engaged with this form of feedback. Employing a quasi-experimental sequential explanatory mixed-methods design, 40 undergraduate EFL students at the University of Oran were assigned to two intact groups: an experimental group (n = 20) receiving AI-mediated corrective feedback and a comparison group (n = 20) receiving teacher-mediated corrective feedback across three writing tasks. Writing accuracy was measured through pre- and post-intervention tasks scored as errors per 100 words. Both groups improved significantly, but the experimental group showed a significantly larger gain than the comparison group (ANCOVA-adjusted difference ≈ 1.87 errors per 100 words, p < .001), while also showing considerably greater variability across individual learners (SD = 1.98 vs. 0.43). Semi-structured interviews with eight students revealed that this variability was closely tied to depth of engagement: students who actively questioned and reflected on AI corrections showed the strongest gains, while passive or uncritical use was associated with minimal or negative change a pattern consistent with the Noticing Hypothesis. The study contributes context-specific empirical evidence from an underexplored North African EFL setting, indicating that AI-mediated feedback can produce substantial, though learner-engagement-dependent, accuracy gains, with implications for feedback-literacy training in higher education writing instruction.
Keywords: AI-Mediated Corrective Feedback; Chatgpt; Writing Accuracy; EFL; Mixed Methods; Algerian Higher Education