Exploring the impact of AI-powered speaking tasks on EFL learners' speaking performance and anxiety: An activity theory study

This Mixed methods study explored the effectiveness of artificial intelligence (AI) interactive speaking tasks in fostering English as a foreign language (EFL) learners' speaking performance and mitigating speaking anxiety through activity theory. Participants, including sixty-four Iranian EFL students, were randomly assigned to the experimental and control groups. While the control group participated in interactive speaking tasks in regular classes, the experimental group performed interactive tasks using SmallTalk2Me. Two tests were administered before and after treatment to measure the participants' speaking performance and anxiety: the IELTS speaking test and a modified form of the Foreign Language Classroom (Speaking) Anxiety Questionnaire (Öztürk & Gürbüz, 2014). Quantitative data were analyzed using Analysis of Covariance (ANCOVA), which reported a significant increase in speaking performance and a meaningful decrease in speaking anxiety in the experimental group. Fifteen learners from the experimental group were interviewed to collect qualitative data. The quantitative results hinted at the shared agency, perceiving a language learner as an augmented subject receiving augmented capabilities. Three primary themes were identified: learner-centered benefits, a supportive learning environment, and associated challenges and limitations. While qualitative findings advocate for self-regulated learning, enhanced learner engagement and motivation, and the significance of constructive feedback, they highlight concerns regarding the excessive reliance on AI, which may undermine autonomy and critical evaluation of alternative options due to automation. The findings indicate a potential trade-off between automation and autonomy in the speaking proficiency of EFL learners. • This study explored the impact of AI tasks on fostering learners' speaking skill. • This study investigated the role of AI tasks in mitigating speaking anxiety. • Activity theory, considering learning as dynamic was used as conceptual framework. • The findings indicate a potential trade-off between automation and autonomy. • Learners' shared agency was cultivated to make them augmented subjects.