Who Gets Heard? Rethinking Fairness in AI for Music Systems

TL;DR

This work tackles cultural representation biases in AI-generated music and their disparate impacts on creators, distributors, listeners, teachers, and students. It analyzes how biased data and design choices shape outputs and stakeholder experiences, drawing on unstructured interviews to ground the discussion in real-world practices. The paper offers concrete interventions at dataset, model, and interface levels, including datasheets, entropy-based diversity auditing, traceability, opt-out mechanisms, and governance considerations, to improve transparency and accountability. It further outlines future research directions such as evaluating fairness in generated music, ensuring credit and consent for communities, extending symbolic representations beyond Western paradigms, and language-aware metadata to mitigate prompts and labeling biases.

Abstract

In recent years, the music research community has examined risks of AI models for music, with generative AI models in particular, raised concerns about copyright, deepfakes, and transparency. In our work, we raise concerns about cultural and genre biases in AI for music systems (music-AI systems) which affect stakeholders including creators, distributors, and listeners shaping representation in AI for music. These biases can misrepresent marginalized traditions, especially from the Global South, producing inauthentic outputs (e.g., distorted ragas) that reduces creators' trust on these systems. Such harms risk reinforcing biases, limiting creativity, and contributing to cultural erasure. To address this, we offer recommendations at dataset, model and interface level in music-AI systems.