Speaker
Description
Speaking anxiety continues to constrain learners' participation and oral performance in English-as-a-Foreign-Language (EFL) classrooms. Although recent work has examined the use of artificial intelligence (AI) in language learning, its role in addressing affective factors remains insufficiently investigated. This study examines the effects of AI-mediated low-stakes practice on speaking anxiety among Vietnamese university students. A quasi-experimental design was implemented with two intact classes of 64 second-year English-majored students at a Vietnamese university. The experimental group engaged in AI-supported speaking activities, including guided role-plays and repeated practice using conversational AI tools over a six-week period, while the control group followed conventional speaking instruction. Data were collected through pre- and post-tests using a speaking anxiety questionnaire adapted from the Foreign Language Classroom Anxiety Scale (FLCAS) and an analytic speaking rubric. The results indicate that students in the experimental group reported significantly lower levels of speaking anxiety and higher levels of confidence compared to those in the control group. Student feedback further suggests that AI-mediated practice enabled repeated, low-pressure exposure and reduced concerns about negative evaluation. The findings provide empirical support for the integration of AI-mediated low-stakes activities in speaking instruction as a means of addressing affective constraints and supporting learner participation in EFL contexts.
Keywords: AI-mediated learning, speaking anxiety, low-stakes practice, EFL classrooms, quasi-experimental design
