Research on Correction Behaviors and Mental Efforts When Using ChatGPT’s Feedback for Revision

Although some studies have explored the effectiveness of using ChatGPT as an automatic writing feedback tool, there has been little research on learners’ behaviors and cognitive load when reacting to the feedback. This study investigated how college-level EFL learners employed feedback generated by ChatGPT for revision and their mental effort when responding to different types of feedback. Thirty-two first-year English majors in an intermediate English composition course participated in this study. They were asked to produce an opinion paragraph. ChatGPT was used to provide feedback on global aspects, including content, coherence, and organization, as well as local aspects such as grammar and vocabulary. Data were collected from their first drafts, final revision, and a questionnaire measuring the mental effort, feedback clarity, and the usefulness of feedback. Using Chi-square tests comparing corrective behaviors across different aspects of feedback, the results showed that students revised their writing significantly more often when addressing local feedback compared to global feedback. Moreover, it was found that clear feedback led to more corrections. On the contrary, general feedback reduced students’ willingness to revise. The paired sample t-tests revealed that the participants reported investing significantly more mental effort when processing global feedback, which may explain why they revised the global issues fewer than the local errors. This study provided novel insights into how students interact with ChatGPT’s automated written corrective feedback, highlighting the need to guide or support students in leveraging such a technology.