YOUNG EFL LEARNERS’ INTERACTION WITH AND PERCEPTIONS OF AI CHATBOT AFFORDANCES FOR SPEAKING PRACTICE

Abstract

 This qualitative study investigated how young EFL learners negotiate meaning during interaction with ChatGPT voice mode and how they perceive its affordances for speaking practice. Ten Vietnamese learners aged 8–12 participated in speaking sessions with ChatGPT followed by semi-structured interviews. Learner–ChatGPT interaction logs were analysed using Long’s Interaction Hypothesis, focusing on clarification requests, confirmation checks, comprehension checks, repetition, and communication breakdowns. Interview data were examined through a five-dimensional affordance framework encompassing perceptibility, learning valence, compositionality, normativity, and intentionality. The analysis identified 68 negotiation episodes across 454 conversational turns, of which 60 (88.2%) were successfully resolved. Clarification requests were the most frequent strategy, followed by confirmation and comprehension checks, while repetition supported interactional continuity. Most learners perceived ChatGPT as useful for sustaining communication and reported lower pressure when requesting assistance, although linguistic limitations occasionally hindered successful resolution. The findings suggest that AI-mediated speaking practice can support negotiation of meaning and interactional problem-solving, while its affordances depend on the interplay between technological resources, learners’ linguistic abilities, perceptions, and active engagement.