Exploring student engagement with ChatGPT feedback: A classroom study of JFL learners

Abstract

Generative AI is increasingly used in L2 writing, yet little is known about its use by beginners and learners of languages other than English. This study examined how four beginner learners of Japanese engaged affectively, cognitively, and behaviorally with ChatGPT feedback while writing. Data were collected through interviews, stimulated recalls, screen recordings, and reflective journals. The data analysis also included error types, prompt types, and the accuracy of ChatGPT output. Affectively, participants valued English translations but reported negative emotions due to inconsistent, self‐contradictory explanations, low trust, and feedback beyond their proficiency. Cognitively, orthographic accessibility reduced attention to feedback, though learners still evaluated suggestions critically. Behaviorally, participants frequently verified ChatGPT output with external resources and used ChatGPT mainly for proofreading and word lookup. Findings highlight how proficiency and orthography shape engagement with AI feedback. This study underscores the importance of teaching AI‐use strategies to elicit level‐appropriate feedback for beginner learners.