Explainable AI through the Lens of Material Agency: Enabling Musical Interface Design with Neural Audio Models

Explainability serves creative exploration, not just transparency: Rather than focusing solely on how AI models make decisions, the paper proposes "material explainability"—making AI models understandable enough for artists and designers to creatively experiment with them. This shifts explainability from a compliance tool to an enabler of sustained creative practice.

Hands-on learning through "playing" with AI is more effective than technical documentation: The research emphasizes that artists learn AI model properties best through direct experimentation and exploration ("at play") rather than traditional explanations. This suggests that for non-technical users, interactive engagement with AI systems may be more valuable than detailed technical specifications.

Community resources and repositories democratize AI access: The case study demonstrates that building shared collections of resources and design examples for neural audio models enables broader participation in AI-based creative work. This practical approach to accessibility suggests that responsible AI adoption benefits from community-driven knowledge sharing, not just top-down guidance.

AI explainability should be tailored to user needs and contexts: The paper's focus on musical interface design shows that explainability requirements differ significantly across domains and user groups. For artists and makers, explainability means understanding creative possibilities; this differs from what explainability means for other stakeholders, highlighting the need for context-specific approaches to responsible AI.