Speaker
Description
The study compares how advanced English-major students apply feedback generated by ChatGPT and Microsoft Copilot to improve language use in college writing essays. It aims to compare the differences across four aspects of language use: lexical accuracy, grammatical accuracy, vocabulary range, and sentence complexity after using ChatGPT and Microsoft Copilot feedback, as well as to explore students' engagement with both tools' feedback during the revision process. A sequential mixed-methods design is adopted. Data are collected from 31 English-major students through an integrated reading-into-writing task, yielding a total of 93 essays, and from 10 participants through semi-structured interviews. Quantitative data were analyzed using one-way repeated measures ANOVA, while qualitative data were analyzed using thematic analysis to identify patterns in students’ engagement with AI-generated feedback. The findings indicate significant differences in three out of four language aspects after revision using ChatGPT and Microsoft Copilot feedback, except for sentence complexity. However, Microsoft Copilot post-feedback version shows a stronger difference in terms of lexical accuracy. Regarding students’ engagement, they generally perceive both tools as clear and useful for revision. ChatGPT is valued for providing broader and more varied language suggestions but is sometimes cognitively demanding. In contrast, Microsoft Copilot feedback is seen as more concise, focus, and easier to apply. Students also demonstrate active, selective, and context-dependent feedback uptake with both AI tools. The study offers pedagogical implications not only for students and teachers but also for institutions, curriculum designers, and educational technology developers.
