Human, AI-generated, and hybrid corrective feedback in L2 argumentative writing

This quasi-experimental study investigated the differential effects of human teacher feedback (HTF), AI-generated feedback (AIGF), and hybrid feedback (HF) on learner engagement and argumentative writing quality among 120 undergraduate ESL students across an eight-week intervention. Engagement (behavioural, cognitive, emotional) was measured via a purpose-built questionnaire and revision-tracking logs; writing quality was scored on a five-dimension analytic rubric. Repeated-measures ANOVA and multiple regression showed that HF was associated with the greatest improvement in writing quality (η²p = .24), HTF with the highest proportion of higher-order revisions, and AIGF with the highest revision frequency. Hybrid feedback condition membership was the strongest predictor of writing-quality improvement, followed by cognitive engagement. These findings have implications for AI-integrated writing instruction in higher education.