The Effects of AI-Powered Automated Written Corrective Feedback on Young EFL Learners’ Writing Performance and Self-Efficacy

Modern technologies, particularly Artificial Intelligence (AI), have transformed the nature, content, and methods of language teaching and learning, including writing instruction, feedback, and evaluation. Given the affordances of AI in automated written corrective feedback (AWCF), it is important to delve into both pedagogical outcomes and learners’ actual experiences and feelings, such as writing self-efficacy (WSE), when exposed to these new possibilities. In the present embedded mixed-methods study, 66 teenage English as a foreign language (EFL) learners, assigned to experimental (n = 34) and control (n = 32) groups, received AWCF via Grammarly and WCF from the instructor, respectively. Both groups completed a self-efficacy questionnaire and six writing tasks, over two months. ANCOVA, MANCOVA, and Quade tests revealed that, while both feedback types improved writing performance, AWCF exerted a significantly greater impact. No significant difference, however, was found regarding WSE. An aggregate of qualitative data from introspective think-aloud protocols and focus group, intended to grasp both real-time engagement and post-intervention insights, revealed mostly favorable attitudes towards the experience. However, learners’ comments pertaining to their sense of self-efficacy displayed more complex patterns marked with mixed feelings. These findings hope to provide worthwhile implications for EFL educators, learners, and educational technologists who are seeking to improve autonomous learning experiences with evolving technologies.