Socio-Affective Strategies in AI-Assisted Speaking: Use, Benefits, and Reflections among Malaysian ESL University Students

Abstract

The growing access to artificial intelligence (AI) speaking tools provides Malaysian tertiary ESL students with opportunities to practise spoken English beyond traditional classroom settings. Despite the growing interest in AI-assisted language learning, limited research has examined how students use socio-affective strategies during AI-assisted speaking practice. This study investigated students’ use of socio-affective strategies, the perceived benefits associated with these strategies, and their reflections on the role of AI speaking tools in supporting speaking development. A mixed methods design was employed involving 165 Malaysian tertiary ESL students with prior experience in using AI speaking tools. Quantitative and qualitative data were collected through an adapted questionnaire based on Oxford’s (1990) Strategy Inventory for Language Learning (SILL). Descriptive statistics, Pearson correlation, multiple regression, and thematic analyses were conducted. The findings revealed that students used both social and affective strategies during AI-assisted speaking practice with affective strategies being used slightly more frequently. Students perceived socio-affective strategies as beneficial for enhancing enjoyment, confidence, fluency, and error management. Affective strategies emerged as the only significant predictor while both strategy dimensions were positively associated with perceived benefits. This finding suggests that the affective dimension may play an important role in shaping students’ perceived benefits within AI-assisted speaking environments. Qualitative findings further indicated that AI features such as feedback, role-play, progress tracking, and avatars supported the use of socio-affective strategies although challenges related to pronunciation, technical issues, and limited human interaction were also reported. The study extends socio-affective strategy research into AI-assisted speaking contexts and offers pedagogical insights for integrating AI-supported speaking activities in Malaysian higher education. More broadly, the findings suggest that AI-assisted speaking environments may represent a distinguished context in which established language learning strategies are implemented through both technological affordances and learner self-regulation.