Effects of ChatGPT on revision practices: Improving learner autonomy and critical engagement―A comparison with DDL practices

This study explores the impact of ChatGPT-3.5 on the academic writing development of Japanese university students in an English writing course. The study is framed through a comparative lens with data-driven learning (DDL) research, aiming to evaluate how insights from DDL can help interpret the outcomes of AI-assisted instruction. Unlike traditional DDL approaches, which emphasize discovery learning through concordance lines, ChatGPT offers immediate, fluent, context-sensitive feedback with minimal cognitive cost. Over a 15-week period, students engaged in iterative writing cycles involving drafting, teacher feedback, and revision using ChatGPT. Results showed improved grammatical accuracy, lexical variety, and rhetorical clarity, along with increased learner confidence and a greater reflective stance toward AI feedback. These findings are discussed in relation to DDL studies that have also documented gains in learner autonomy, metalinguistic awareness, and writing quality based on corpus use. This study highlights how AI-assisted revision and DDL share complementary goals in promoting informed, independent writing. The findings suggest that integrating generative AI with pedagogical principles from DDL can enhance learner agency and writing proficiency in EFL contexts.