|
چکیده
|
Despite the rapid uptake of ChatGPT in English as a foreign language (EFL) writing instruction, existing research has largely examined the general effects of AI access rather than isolating the effects of self-regulated AI use. Studies that jointly examine paragraph-level writing performance and writing self-efficacy within a controlled design remain limited, particularly in the Iranian EFL context. This study employed an explanatory sequential mixed-methods design to investigate whether the self-regulated use of ChatGPT was associated with changes in Iranian EFL learners’ compare-and-contrast and cause-and-effect paragraph-writing performance and writing self-efficacy, while exploring the underlying processes through learners’ experiences. In the quantitative phase, a quasi-experimental pre-test/post-test design was used with 64 Iranian EFL learners in two intact university writing classes (experimental n=33n = 33n=33; control n=31n = 31n=31). Over a 12-week intervention, the experimental group received explicit training in self-regulated strategies for consulting ChatGPT during paragraph writing, whereas the control group received conventional instruction. Writing performance was assessed through analytic rubrics, and self-efficacy was measured using the L2 Learners’ Writing Self-Efficacy Scale (LWSE). In the qualitative phase, semi-structured interviews and reflective learning logs were analyzed thematically to explain and contextualize the quantitative findings. Results of separate MANCOVAs indicated that the experimental group significantly outperformed the control group in both paragraph genres across all analytic sub-dimensions. Additionally, independent-samples t-tests revealed a substantial post-intervention advantage in writing self-efficacy for the experimental group (Cohen’s d=2.23d = 2.23d=2.23). Thematic analysis identified three interrelated themes: scaffolded idea generation and reduced writer’s block, greater awareness of paragraph structure and cohesion, and improved accuracy alongside a self-recognized risk of dependence. Overall, the findings suggest that guided, self-regulated engagement with ChatGPT supports EFL paragraph writing and self-efficacy. The study contributes to self-regulated learning and generative AI theory and offers implications for AI-integrated EFL curricula and teacher training.
|