From assistance toward appropriation: An activity-theoretical perspective on ChatGPT-mediated academic writing among university students

Abstract

Generative artificial intelligence has created new possibilities for supporting second language academic writing, but less is known about how learners take up such support within ordinary writing instruction. Drawing on Activity Theory, this qualitative classroom-based case study examined how ChatGPT was used in an Uzbek university academic writing course and what tensions shaped students' movement from AI assistance toward appropriation. The participants were 10 first-year EFL students majoring in English literature and English language teaching at Fergana State University. Across a ten-week course, data were collected through a background questionnaire, baseline writing task, six AI-mediated writing tasks, ChatGPT interaction logs, draft-revision pairs, reflective writing logs, semi-structured interviews, stimulated recall interviews, and teacher field notes. Thematic analysis identified three themes. First, students initially used ChatGPT as a substitute for difficult writing work, especially for paragraph generation, correction, and rewriting, but gradually developed more task-specific uses. Second, ChatGPT created a tension between convenience and learning responsibility, as it helped students produce more fluent paragraphs while sometimes weakening control over meaning, authorship, and revision decisions. Third, teacher mediation, including AI-use rules, prompt modeling, feedback, reflection, and classroom dialogue, helped regulate AI use. The findings suggest that ChatGPT-mediated writing development depends not only on tool access but also on pedagogical mediation that helps learners evaluate, regulate, and explain AI-supported writing decisions.