Learner Use of AI-Generated Feedback for Written Corrective Feedback in L2 Writing: Usefulness,User Proficiency and Attitude

Written corrective feedback (WCF) plays a crucial role in developing L2 writing skills, but its effectiveness is often limited by constraints faced by instructors. The emergence of Artificial Intelligent (AI) language models makes it possible for personalized, on-demand feedback to learners. This study investigates the usefulness of AI-generated WCF in improving learners’ academic writing abilities, their proficiency in and attitude towards utilizing such feedback. 22 first year Chinese university students participated in a 6-week experiment consisted of pre/post writing tests and 8 writing practices. In the practices, participants used feedback from available AI models (ChatGPT, Bard, Gemini) to revise their drafts. Quantitative data from pre/posttests and a Likert-scale questionnaire, alongside qualitative responses, were collected and analyzed. An average 0.78-point increase in the post test indicated noticeable improvements in the participants’ writing abilities. Questionnaire responses revealed a positive perception of AI WCF's usefulness in enhancing grammar, vocabulary, coherence, and academic style. While proficient in leveraging AI models for feedback, some participants faced challenges in prompting and interpreting answers. Most displayed a favorable attitude towards adopting AI WCF for writing both in and outside the classroom, despite concerns regarding creativity and academic integrity. The findings support the pedagogical value of incorporating AI technologies to facilitate L2 writing development. However, developing literacy using AI models and addressing ethical concerns remain critical in this quest.