CMU M&T Seminar: The Creativity Cost of Large Language Model (LLM) Ideation in Music Composition скачать в хорошем качестве

CMU M&T Seminar: The Creativity Cost of Large Language Model (LLM) Ideation in Music Composition

This presentation was given as part of the Music and Technology Seminar (Spring 2026) of Carnegie Mellon University. Date: 01/20/2026 Abstract: Generative AI is transforming creative industries, particularly music, and raising questions about artistic practice and creator roles. Text-to-music systems can produce full songs from prompts, yet their effect on human creativity and quality is unclear. This paper analyzes how AI assistance shapes the creative process and outcomes in music. We conduct a randomized controlled trial with musically trained individuals who compose an original 15-second piano melody, assigning them to either an AI-ideation group or a no-AI control. External evaluators of varied musical experience rate each melody on perceived creativity, musicality, and enjoyment. On average, we find that average listeners rate the melodies generated under the treatment condition as less creative. Furthermore, access to the AI changes stylistic choices, which are perceived as less creative by evaluators. About the speaker: Jose Orós is a Ph.D. candidate in Information Systems Management at Carnegie Mellon University's Heinz College. He studies how digitization and technology impact media markets, with a focus on the music industry. He is especially interested in how digital platform policies, such as recommendation algorithms and editorial decisions, influence consumer and creator behavior and drive off-platform outcomes. He is also interested in the opportunities and risks of AI in creative industries.