Affordances and constraints of ChatGPT feedback in ESP writing: Vietnamese medical students’ perspectives

Abstract

This study investigates student attitudes toward AI-generated feedback in second language (L2) writing teaching. Conducted with 55 Vietnamese medical students enrolled in an English for Healthcare Professionals course, the study included one writing task evaluated by ChatGPT-4o. After receiving the feedback, students completed a perception-based survey comprising seven Likert-scale items and two open-ended questions. Descriptive statistics and inductive thematic analysis were employed to analyze student responses. The findings reveal a complex preference pattern. Students appreciated AI feedback for its timeliness, comprehensiveness, and clarity in grammar and vocabulary correction. However, they saw it as emotionally distant, mechanical, and possibly inaccurate or overwhelming. Participants advocated for a hybrid model of feedback that would pair the technical efficacy of AI and the human insight of instructors. These findings contribute to the growing literature on AI in L2 writing and highlight the need for careful integration of AI-generated feedback in discipline-specific language learning contexts.