<![CDATA[Developing interactional competence (IC) remains a significant challenge in Japanese elementary English teacher education, where pre-service teachers tend to prioritize linguistic accuracy over pedagogical facilitation. This pilot study investigates how a generative AI-based virtual dialogue environment—one specifically designed to introduce “friction” through unpredictable and incomplete responses—may reshape pre-service teachers' task awareness. Two university students took part in a two-week intervention using ChatGPT’s voice mode, assuming the role of teachers interacting with an AI “child.” Semi-structured interview data were analyzed using a Grounded Theory Approach (GTA). The analysis reveals a four-stage process: (1) initial difficulty arising from the gap between expectations and reality, (2) a shift in awareness from self-oriented linguistic anxiety toward learner-focused facilitation, (3) the concretization of perceived classroom conflicts, and (4) an emerging desire for pedagogical support. A central finding is that AI-mediated “friction” serves as a productive catalyst for professional learning, destabilizing existing frames of reference and prompting participants to redefine IC as the co-construction of meaning. The study carries several implications for teacher education and AI design. It proposes the principle of “intentional imperfection” in AI behavior, suggesting that rather than providing flawless models, AI for teacher training should generate manageable interactional trouble to elicit pedagogical judgment. The findings also highlight the necessity of embedding such simulations within a scaffolded framework that provides diagnostic feedback and links virtual practice to real-world professional vision. Together, these contributions offer a conceptual foundation for utilizing AI to prepare teachers for the interactional complexities of the language classroom.]]>
Copyrights © 2026
