NEUTRAL BUT HELPFUL? INVESTIGATING THE USE OF CHATGPT FOR NON-JUDGMENTAL FEEDBACK IN ACADEMIC WRITING

Abstract

With increasing inclusion in learning environments, the value of generative AI in facilitating formative feedback in academic writing needs critical exploration. This study examines how non-judgmental feedback in ChatGPT differs from feedback provided by teachers and examines students’ reception of both. Informed by Feedback Literacy Theory, the research integrates a qualitative comparative case study with a total of 30 university students majoring in English Language Studies in a public university in Ho Chi Minh City, Vietnam. Each student wrote two pieces of academic writing, marked independently by experienced instructors and ChatGPT with a common rubric. Feedback texts were thematically analyzed based on tone, area, and depth. Six focus group interviews were also conducted to gain insight into students’ reception and engagement with the feedback. Results show, in contrast with teacher feedback, which inclined towards a mixture of directive comment with evaluative judgment, ChatGPT exhibited a consistent facilitative, non-judgmental tone. Students liked the polite, emotionally neutral tone in ChatGPT feedback, with comments citing it as less threatening, more encouraging for revision. However, a lack of depth, in addition to a lack of task-specificity, in AI feedback were also reported. The research suggests the value in incorporating ChatGPT as a low-stakes, non-judgmental feedback tool in support of teachers’ feedback, instilling in students feedback literacy, especially in L2 writers. Implications for writing instruction include strategic integration towards an enhanced self-regulatory learning condition, with a caution against over-reliance on program-based feedback. The study calls for future longitudinal research investigating the developing status of AI in assisting academic writing development.