Human–AI Complementarity in Secondary L2 Writing: A Within-Subject Investigation of Feedback Modality, Self-Regulation, and Revision Processes
Description
The integration of Artificial Intelligence (AI) into language education has transformed feedback practices in second language (L2) writing. While prior research has demonstrated the effectiveness of AI-assisted feedback in improving linguistic accuracy, empirical evidence in secondary ESL contexts remains limited, particularly regarding how AI reshapes learner engagement and revision processes. This study investigates the impact of sequential feedback modalities—teacher-mediated followed by AI-assisted feedback (via ChatGPT free version)—on formal letter writing among secondary-level ESL learners (N = 39) enrolled in Grade 11 under the Central Board of Secondary Education (CBSE) curriculum in India. Employing a within-subject quasi-experimental design, all participants experienced both feedback conditions across structured revision cycles.
Files
Steps to reproduce
This study investigates the impact of sequential feedback modalities—teacher-mediated followed by AI-assisted feedback (via ChatGPT free version)—on formal letter writing among secondary-level ESL learners (N = 39) enrolled in Grade 11 under the Central Board of Secondary Education (CBSE) curriculum in India. Employing a within-subject quasi-experimental design, all participants experienced both feedback conditions across structured revision cycles. Writing performance was assessed using an analytic rubric across five domains, and self-regulation and AI perception were measured using validated scales (Cronbach's α = .78 and .82, respectively). Paired-samples t-tests revealed a statistically significant improvement in writing performance following AI-assisted feedback compared to teacher-mediated feedback, t(38) = 3.21, p = .003, with a moderate effect size (d = 0.51). Domain-specific analysis indicated stronger gains in lower-order features (grammar and mechanics), while improvements in higher-order features (content and organization) were comparatively modest. Regression analysis demonstrated that self-regulation significantly predicted writing improvement (β = .42, p = .008), accounting for 19% of the variance. Additionally, AI perception was positively correlated with both self-regulation and writing outcomes (r ≈ .50). Interpreted through Zimmerman's (2000) self-regulated learning framework and Vygotskian sociocultural perspectives on mediated feedback, the findings suggest that AI functions as a mediational tool that enhances revision efficiency and supports self-regulated writing processes rather than replacing teacher scaffolding. The study contributes to the emerging discourse on human–AI complementarity by demonstrating how the sequential integration of teacher and AI feedback can optimize writing development in secondary ESL classrooms.
Institutions
- Amrita Vishwa VidyapeethamTamil Nadu, Coimbatore