Learning by Teaching: Enhancing Music Learning Through LLM-Based Teachable Agents

This study presents Chat Melody, an LLM-based multimodal teachable agent designed to enhance music theory learning through the Learning-by-Teaching (LBT) approach. By simulating a novice learner, Chat Melody engages students in structured dialogue that prompts explanation, error identification, and conceptual refinement. The system integrates a multimodal interface combining visual notation, auditory feedback, and AI-generated explanations. LBT is operationalized through a structured prompting strategy based on the help-seeking model, guiding progressive reasoning instead of offering direct answers. In a quasi-experimental study with 28 students, Chat Melody was compared to traditional teacher-guided instruction. Results showed that students using the teachable agent demonstrated greater improvements in knowledge construction. Additionally, responses to the Situational Motivation Scale (SIMS) indicated higher intrinsic motivation and perceived value in music learning. These findings underscore the potential of LLM-based teachable agents to support meaningful learning and cognitive engagement in AI-assisted music education.