As generative AI continues to be adopted in educational settings, new possibilities for teaching and learning are emerging across various subjects. In music education, interest has grown in whether AI-assisted composition lessons can supplement the ch...
As generative AI continues to be adopted in educational settings, new possibilities for teaching and learning are emerging across various subjects. In music education, interest has grown in whether AI-assisted composition lessons can supplement the challenges of traditional creative instruction and enhance students’ creative capabilities. However, in the field of education, generative AI is still being used as a tool or is limited to one-time experiential activities. There is a lack of instructional design and empirical analysis aimed at systematically strengthening students’ subject knowledge, compositional ability, and self-directed learning. Particularly in an era where AI-generated artistic works often surpass human creations, it is critical to utilize AI in a way that preserves the essence of the discipline while achieving meaningful educational goals. While some studies have explored generative AI in creative learning, most have focused only on prompting and generating outputs, leaving it unconfirmed whether students' subject knowledge and creative competencies actually improved through the lessons.
Against this backdrop, this study aims to analyze the effects of a generative AI-based music composition class on middle school students’ music knowledge, compositional competence, and musical self-efficacy. An experimental study was conducted with 80 second-year middle school students, who participated in a three-session music composition class using either a knowledge-based AI tool (MusiaOne) or a prompt-based AI tool (Suno). The lessons included fundamental music theory for single-phrase composition, hands-on composition using AI tools, and a discussion-based session introducing alternative AI types. Pre- and post-surveys and in-depth interviews were used to compare and analyze the effects. Survey items were validated by experts using CVI and IRA methods, and statistical analyses included paired and independent samples t-tests.
The results showed statistically significant improvements in music knowledge, compositional competence, and musical self-efficacy in both groups. In particular, the group using MusiaOne demonstrated higher self-directed engagement and understanding of the composition process, having directly applied music theory during the activity. In contrast, while the Suno group found the instant generation of music interesting, they showed a lower degree of internalization of compositional principles and musical structure. In-depth interviews confirmed that students felt greater satisfaction and accomplishment when independently shaping their musical ideas, suggesting that appropriate collaboration with AI constitutes meaningful learning.
This study provides empirical evidence that generative AI can serve not merely as a technological tool but as an instructional strategy for achieving curriculum goals. In particular, AI-integrated lesson designs that emphasize understanding of musical elements and growth in creative competencies can offer practical direction for developing integrated AI curricula in various arts and humanities subjects. It is hoped that this music composition lesson model will be actively researched and shared in schools, allowing generative AI to serve as a partner in fostering creativity and self-directed learning.