Students' Voices on the Implementation of Gen-AI Feedback in EFL Writing Classroom

Abstract

The proliferation of generative artificial intelligence (Gen-AI) in English as a Foreign Language (EFL) writing instruction has sparked considerable scholarly interest, yet the subjective experiences and voices of learners who engage with AI-generated feedback remain insufficiently explored within the extant literature. This qualitative study investigates students' voices concerning the implementation of Gen-AI feedback in an EFL writing classroom, specifically examining their perceptions, patterns of engagement, development of feedback literacy, and the inherent tensions between learner autonomy and technological dependency. Adopting a qualitative research design, data were collected through four focus group discussions involving 30 purposively selected undergraduate students enrolled in an EFL writing course at a private university in Surakarta. Thematic analysis of the transcribed discussions, conducted in accordance with Braun and Clarke's (2006) framework, yielded five interrelated themes: perceived usefulness of Gen-AI feedback, patterns of feedback engagement, development of feedback literacy, tension between autonomy and dependency, and emotional and ethical responses. The findings reveal that students perceived Gen-AI feedback as highly useful for enhancing writing quality, revision efficiency, and idea generation; however, they did not passively accept all AI-generated suggestions but rather demonstrated selective engagement, critically evaluating feedback based on its relevance to higher-order writing concerns, including content and organization. Additionally, participants reported developing feedback literacy over time, evidenced by increased awareness of their writing weaknesses and improved capacity to interpret and strategically utilize AI feedback. Nevertheless, uneven development across participants indicated that feedback literacy does not emerge automatically through mere exposure to Gen-AI tools. A critical tension also emerged between learner autonomy and dependency, with several students expressing over-reliance on AI for revision tasks. Emotional and ethical responses included heightened confidence alongside scepticism regarding AI accuracy and concerns pertaining to academic integrity. The findings suggest that the pedagogical effectiveness of Gen-AI feedback is not determined solely by its technological affordances but is significantly mediated by learners' engagement, critical awareness, and feedback literacy. This study contributes to the literature by demonstrating that understanding students' voices is crucial for developing context-sensitive and pedagogically sound AI integration strategies, with implications for educators and curriculum designers seeking to implement Gen-AI feedback in effective, ethical, and learner-centred ways.