AI-Assisted Language Learning, Self-Efficacy, and Speaking Anxiety in ESL Contexts: A Literature Review

Abstract

In second language acquisition, oral anxiety significantly limits learners' willingness to communicate and academic achievement. Though AI conversational systems offer low-stress oral practice, the psychological mechanisms underlying anxiety reduction—and how individual differences such as personality traits moderate these effects—remain underexplored. This review brings together research on artificial intelligence-assisted language learning, self-efficacy and oral anxiety in the ESL (English as a Second Language) environment and consciously focuses on introverted learners. The thematic comprehensive analysis of the literature shows that the artificial intelligence-mediated environment usually reduces situational oral anxiety and enhances learners' self-confidence by providing non-judgmental interactions and repeated drills. Based on Bandura's theory of self-efficacy, the analysis further points out that the improvement of perception ability is the core intermediary mechanism that connects AI interaction and anxiety reduction. However, the evidence in different intervention designs is far from consistent, and introverted learners - who usually experience severe oral anxiety with an unbalanced degree in traditional classrooms - are still a group with an obvious lack of research. This review concludes that while AI shows promise in alleviating oral anxiety, its effectiveness depends on learner characteristics and pedagogical context; future research should adopt learner-centered designs to address these gaps.