DeformTune: A Deformable XAI Music Prototype for Non-Musicians

Physical interfaces can make AI music creation more accessible: DeformTune's tactile, deformable interface allowed non-musicians to engage with AI music generation without requiring technical knowledge or musical training, suggesting that embodied interaction design is a viable approach for democratizing AI creative tools.

Current AI music systems lack clear explanations of how user inputs affect outputs: The study identified "unclear control mappings" as a recurring challenge, indicating that even intuitive interfaces fail if users don't understand the relationship between their actions and the AI's musical decisions—a critical gap in explainability.

Non-musicians need guided, progressive support to use AI music tools effectively: Participants required ongoing guidance throughout their use of the system, suggesting that one-time tutorials are insufficient; AI music interfaces should incorporate scaffolded learning and real-time support to help novice users build confidence.

Multimodal feedback enhances AI explainability: The paper identifies multimodal feedback (combining visual, auditory, and tactile information) as a design opportunity that could help users better understand how the AI interprets their inputs and generates musical outputs.

Limited expressive range undermines user agency: Participants encountered constraints in what musical variations they could create, highlighting that explainability alone is insufficient—AI systems must also provide users with meaningful creative control to feel genuinely empowering.