Speaker
Description
Abstract
Generative AI is increasingly embedded in academic writing, yet existing research has focused more on writing outcomes and learner perceptions than on how learners negotiate control during AI-mediated revision. Grounded in a sociocultural understanding of agency as mediated, relational, and context-dependent, this qualitative multiple-case study explored how eight Vietnamese university EFL students revised two academic essays through two AI-assisted revision cycles over six weeks. Data comprised successive drafts, AI interaction logs, reflective notes, and stimulated-recall interviews. Thematic analysis of interactional and reflective data was integrated with revision analysis across drafts. Findings indicated that agency was task-contingent and shifted within and across cases. At some points, students went beyond simply copying AI-generated text, iteratively prompting AI, comparing alternatives, challenging inaccurate suggestions, and selectively adapting output to preserve intended meanings and rhetorical purposes. At other points, they accepted polished reformulations with limited evaluation or focused primarily on surface-level correction. These shifts were associated with linguistic confidence, perceived task difficulty, and calibrated trust in AI. The study conceptualises AI-assisted revision as a process in which decision-making control is repeatedly enacted and redistributed. For TESOL pedagogy, the findings highlight the need to develop learners’ evaluative judgement, purposeful prompting, and accountable revision practices rather than treating AI literacy as tool use alone.
Keywords: generative artificial intelligence, learner agency, academic writing revision, AI-mediated feedback, Vietnamese higher education
