AI-Generated Feedback and Second Language Writing Development: A Longitudinal Investigation of Learner Autonomy, Accuracy, and Affective Response in Higher Education Contexts

Abstract

This longitudinal mixed-methods study investigates the long-term impact of AI-generated feedback tools such as ChatGPT and Grammarly on undergraduate EFL/ESL learners’ academic writing development over three academic years[1][2]. Addressing gaps in existing automated written corrective feedback (AWCF) research, the study examines three key dimensions: linguistic development, learner autonomy, and affective outcomes. Approximately 80 university students will be tracked across six semesters through quantitative measures of syntactic complexity, lexical diversity, error frequency, writing anxiety, self-efficacy, and foreign language enjoyment, alongside qualitative data from interviews, reflective journals, stimulated recall sessions, and screen-capture logs. The study aims to provide deeper insight into how learners engage with AI-generated feedback and how it influences their revision practices, long-term interlanguage development, and emotional responses to writing. Ultimately, the research seeks to contribute to theories of AI-mediated second language acquisition and offer practical recommendations for integrating AI and human feedback effectively in higher education L2 writing instruction.