Speaker
Description
Abstract
Against the backdrop of the “IELTS craze” in Vietnam and the rapid integration of generative AI into language education, this study examines Vietnamese EFL preservice teachers’ perceptions of the validity of using ChatGPT to assess IELTS Writing Task 2. Guided by Kane’s (2013) argument-based validation framework and Hattie and Timperley’s (2007) model of effective feedback, two research questions are addressed: how preservice teachers perceive the validity of ChatGPT-generated scores, and how they perceive the validity of ChatGPT-generated feedback. A qualitative design was adopted, and semi-structured interviews were conducted with ten fourth-year students majoring in English Language Teaching Education at a university in Hanoi. All participants had completed a language testing course, had at least two years of IELTS Writing Task 2 tutoring experience, and had used ChatGPT in their assessment practice. Thematic analysis revealed that ChatGPT was perceived as reasonably accurate for Lexical Resource and Grammatical Range and Accuracy, but less reliable for Task Response and Coherence and Cohesion, where it tended to inflate scores and overlook weaknesses in argumentation. Participants also noted prompt sensitivity, sycophantic score adjustment, and cross-session memory effects. Feedback was valued for speed but criticised as generic, mechanically worded, and misaligned with learner proficiency. These findings have significant implications for TESOL, particularly in developing AI literacy among preservice teachers and informing responsible integration of AI-based tools in language teacher education.
Keywords: ChatGPT, IELTS writing assessment, preservice teachers, assessment validity, AI literacy
