Current Issue
Volume 21, Issue 2, July 2026
The Perceived Boundaries of Acceptable AI Use in EFL Writing: Divergent Views from Students and Teachers
Christopher Andrews, Westgate Corporation
Abstract
This cross-sectional study compared the views of 280 students and 32 teachers at a Japanese university to understand attitudes towards AI use in first- and second-year English EFL writing tasks. Eleven items rated on a four-point academic acceptability scale (1= Completely acceptable, 4= Completely unacceptable) included direct content generation, editing, brainstorming, and translation. Group differences were analyzed using Welch's t-tests and Mann-Whitney U tests, with effect sizes reported. The findings reveal substantial divergence between students and teachers, particularly regarding AI translation and AI-generated or heavily AI-edited text, which teachers judged significantly less acceptable. In contrast, lighter, process-oriented uses of AI showed greater agreement. These differing views as well as high levels of intra-group disagreement, reveal the absence of a consensus surrounding acceptable AI use in EFL writing, which suggests a need for clearer institutional guidelines, improved AI literacy. By highlighting divergent perceptions of acceptable AI use among students and teachers, this study contributes to the ongoing debate around acceptable AI use in EFL writing.
From Form-Focused Corrections to Learner-Initiated Tasks: A Conversation Analysis of ChatGPT in EFL Speaking Practice
Huang Mengqi, Yamaguchi University, Hu Fei, Fudan University
Abstract
This study explores the interactional dynamics of ChatGPT-mediated English speaking practice among Chinese university EFL learners, drawing on Conversation Analysis (CA) to examine how form-focused corrections and learner-initiated task negotiations are sequentially organized in human-AI dialogues. Ten undergraduates participated in role-play tasks with ChatGPT, designed to elicit spontaneous speech and followed by AI-generated feedback. The analysis identifies recurrent practices in which ChatGPT orients primarily to form-focused corrections, while learners display orientations to meaning-making, topical progression, and conversational coherence. Transcript excerpts further show that learners frequently initiate new tasks, modify prompts, or reframe scenarios to sustain interactional flow, thereby influencing the trajectory of the dialogue. While AI feedback provides immediacy and adaptability in task negotiation, it offers limited affective support and narrative guidance compared to teacher-led interactions. These findings shed light on how generative AI reshapes turn-taking, repair practices, and task management in EFL contexts, and suggest pedagogical implications for integrating AI into speaking practice. The study argues for prompt design and post-interaction scaffolding to complement AI's corrective orientation with more contextual and affective resources.