Modes of Engagement with AI-Generated Feedback: A Case Study of Three Pre-Service EFL Teachers

Abstract

Generative AI (GenAI) feedback tools are increasingly used to support formative writing instruction, yet most research reports feedback effects in aggregate, offering little insight into how individual learners actually engage with AI-generated feedback. This exploratory case study examines three pre-service EFL teachers' engagement with feedback from CyberScholar, a rubric-aligned AI feedback platform, during revision of a short reflective essay on AI in English language education. Drawing on a larger pilot dataset, three contrasting cases were purposively selected to illustrate distinct modes of engagement: regenerative rewriting, in which the participant produced an almost entirely new text in a markedly more formal register; minimal, selective engagement, in which only a narrowly targeted mechanical issue was addressed while conceptual feedback was left unaddressed; and partial, uneven engagement, in which some rubric criteria were revised while others remained unchanged. Comparing participants' draft-to-draft changes, the AI-generated feedback they received, and the resulting rating changes, the study develops a preliminary typology of engagement modes rather than a claim about writing-quality improvement. Findings suggest that the same feedback instrument can prompt qualitatively different revision behaviors depending on how learners interpret and act on it, with implications for how AI feedback tools are designed and how teacher educators might scaffold students' critical use of such tools. The study contributes a fine-grained, learner-centered perspective that complements aggregate analyses of AI feedback effectiveness.