Teacher-managed generative AI for personalized learning in intermediate Japanese: A classroom-based design study of reading logs and oral information sharing

Abstract

This classroom-based design study examines how a teacher-managed, generative AI-supported learning system can be integrated into intermediate Japanese as a foreign language (JFL) instruction to support personalized learning and oral information sharing. While large language model-based generative AI can generate texts, explanations, questions, and feedback, its value in language teaching depends on how such affordances are aligned with pedagogical goals and classroom interaction. A teacher-managed web-based system was implemented over seven instructional sessions with five third-year Japanese majors. Data included system logs, task scores, timestamps, classroom observations, mini-interviews, and learner-produced presentation scripts. Findings from this exploratory design study suggest that personalization emerged through teacher-controlled prompt revision, learner-regulated pacing and sequencing, and immediate AI-generated feedback. Learning logs, interpreted together with classroom observations and mini-interviews, revealed iterative and expansive pathways of self-regulated learning. Presentation-script analysis showed that learners initially juxtaposed information from different texts, then stabilized information selection through same-text collaborative tasks, and later began to reorganize information under emerging higher-order categories. The study argues that generative AI is pedagogically meaningful not as an autonomous content generator, but as part of a teacher-designed learning ecology that connects individualized preparation, log-based formative assessment, and collaborative classroom use.