How L2 Writers Engage With LLM Feedback
This project studies what second-language writers actually do with feedback generated by large language models: which suggestions they take up, which they resist, and how their texts and revision strategies change over the course of a writing task.
The study uses quantitative ethnography, combining epistemic network analysis (ENA) of writing-process data with close analysis of drafts and revisions, so that patterns of engagement with LLM feedback can be traced rather than only self-reported.
A pilot phase has been completed and was presented at ICQE 2026. The pilot mapped how L2 writers used ChatGPT during a source-based writing task, including how they sought and responded to its feedback, and the main study now isolates that feedback engagement. The main study is currently in preparation.
Tianshi Hao is a Specially Appointed Assistant Professor at Hiroshima University.