AI partner or peer partner? Exploring AI-mediated interaction in EFL pronunciation from a socio-cultural perspective

This sequential explanatory mixed-methods study examines the effects of AI-mediated pronunciation interaction on the pronunciation skills of English as a Foreign Language (EFL) learners, as well as its influence on foreign language anxiety and enjoyment, framed within Vygotsky's sociocultural theory. Learners' attitudes toward AI-mediated pronunciation activities were also qualitatively explored. Participants ( N  = 100) were divided into an experimental group ( N  = 51), who engaged in AI-mediated interactive speaking activities using EAP TALK, and a control group ( N  = 49), who participated in peer-mediated pronunciation tasks via WeChat. Quantitative data were collected through a pronunciation test and an EFL anxiety and enjoyment scale, and analyzed using paired and independent samples t -tests along with ANCOVA. Qualitative data were derived from individual reflective reports and analyzed thematically. Results indicated that both groups improved their pronunciation proficiency and enjoyment, but the experimental group showed significantly greater gains in these areas. The sociocultural analysis revealed that AI-mediated tools served as digitally embedded more knowledgeable others, offering adaptive scaffolding aligned with Vygotsky's Zone of Proximal Development (ZPD). Moreover, peer-mediated learning also facilitated meaningful linguistic interaction and emotional support, though the quality of feedback varied across groups. While learners generally expressed positive attitudes toward AI-mediated instruction, they also reported challenges related to speech recognition and feedback accuracy. These findings highlight the complementary strengths of AI- and peer-mediated approaches, offering practical implications for the integration of technology-enhanced pronunciation instruction in EFL contexts.