Beyond Classroom Walls: AI-Powered Speaking Practice as Catalyst for Enhanced Self-Efficacy and Communicative Engagement in EFL Learners
Keywords:
AI-Mediated learning, EFL secondary students, speaking self-efficacy, willingness to communicateAbstract
The integration of Artificial Intelligence (AI) in language education has extended speaking practice beyond the constraints of traditional classroom environments. However, limited research has explored how AI-mediated interaction influences learners’ psychological dimensions, particularly speaking self-efficacy and willingness to communicate (WTC), in secondary school contexts. Addressing this gap, this study investigates how AI-powered conversational practice using Replika AI functions as a catalyst for enhancing EFL students’ self-efficacy and communicative engagement. This qualitative study involved 15 tenth-grade EFL students from SMAN 19 Tebo, Jambi, Indonesia. Data were collected through classroom observations and semi-structured interviews following a 15-minute AI-mediated speaking activity focused on self-introduction. The data were analyzed using deductive thematic analysis guided by Bandura’s self-efficacy theory and MacIntyre’s WTC model. The findings reveal that AI-mediated interaction creates a low-anxiety, non-judgmental environment that encourages repeated speaking practice. Key factors contributing to this effect include the absence of peer evaluation pressure, learners’ control over pacing and repetition, and reduced self-consciousness due to the lack of face-to-face interaction. These conditions significantly enhanced students’ speaking self-efficacy, particularly their confidence, which in turn fostered greater willingness to communicate. However, the findings also indicate that this increased communicative readiness remains context-dependent and does not fully transfer to real classroom settings. This study contributes to the growing body of research on AI in EFL by highlighting the psychological mechanisms through which AI supports language learning. The findings suggest that AI-powered conversational tools can serve as effective complementary resources to promote learners’ confidence, autonomy, and engagement, particularly in low-proficiency contexts.

