Generative AI vs. Teachers: Feedback Quality, Feedback Uptake, and Revision
Hetian Yu, Qin Xie
Abstract
Generative AI has stormed the education community worldwide. Gen-AI’s potential to assist teachers in providing feedback to students is the focus of the current research. Specifically, the research investigated how similar or different AI-generated feedback was from teacher feedback, and whether there were differences in students’ responses to the two feedback sources. The research adopted a 2x2 counter-balance experimental design (Latin Square), whereby the participants were divided into two groups to write on two topics, receive feedback, and resubmit the revised work. The participants were 60 EFL secondary students and four teachers from a high school in China. All teachers and a Gen-AI bot (ChatGPT) marked and provided feedback on the 240 written samples. The analysis of the 1200 records (240*5) found that the AI-bot’s surface-level feedback was comprehensive, accurate, and similar to that of one teacher. It outperformed the teacher qualitatively and quantitatively in the meaning-level feedback. Overall, teacher feedback received higher student uptake rates, but the difference was small. Students were capable of adopting a variety of strategies to respond to the feedback regardless of its sources. The findings support adopting ChatGPT feedback as a supplement to teacher feedback given to novice EFL writers.