LEVERAGING CHATGPT'S ADVANCED VOICE MODE FOR ENHANCING ORAL-AURAL SKILLS IN HIGHER EDUCATION LANGUAGE LEARNING: A MIXED -METHODS STUDY

Abstract

The growth of sophisticated conversational artificial intelligence (AI) technology has had a large impact on higher education, but there is limited empirical research into its effectiveness for developing oral-aural skills. Although AI for language learning has previously focused on text-based skills, the development of powerful voice interaction technology opens up new opportunities for pedagogy. This paper fills this gap by examining the effect of ChatGPT’s Advanced Voice Mode on the oral proficiency development of EFL learners at a university level. Following a mixed methods explanatory sequential model design, this study quantitatively determined the improvements in learners' oral proficiency over an intervention period of 8 weeks. The next phase was qualitative, where we elicited learners' perspectives and experiences though semi-structured interviews. Quantitative findings indicate that fluency and lexical resource significantly improved, but pronunciation was found not to improve significantly. This discrepancy was promoted through qualitative data, focusing on three main themes: the AI's potential for low-anxiety rehearsal spaces, learners' annoyance with the tool's lack of nuanced feedback about accent phenomena and the difficulty associated with its real-time corrective function. The experimental results indicate that ChatGPT’s voice mode is a powerful tool to help develop learners’ confidence and fluency in conversation, but is best used as an adjunctive instrument with other means of pedagogical support. It cannot take the place of the subtle, phonetic and pragmatic model that teacher- led instruction provides. These results inform a proposed underpinning "Blended Dialogic Model" to incorporate conversational AI in language curriculum, focusing on technology-mediated practice and expert human tutorial working synergistically.