Music Recommender Systems: Call for Papers by ACM TORS for a Special Issue

Music recommendation is an unusually difficult problem. It is not only about matching listeners to songs. It is also a matter of balancing the interests of different stakeholders, whose interests do not always align: users want relevance and discovery, artists and other creators want visibility and fair exposure, and streaming platforms want engagement, retention, and business success.

The call for papers for the Special Issue on Music Recommender Systems in ACM Transactions on Recommender Systems rightly acknowledges this complexity, including multi-stakeholder optimization, fairness, transparency, and the broader societal impact of algorithmic recommendation.

And music recommendation goes well beyond the familiar streaming homepage. We usually think first of Spotify, Apple Music, or YouTube Music. But recommendation matters in many other settings as well. A filmmaker may search for the right piece of music for a scene. A record label may want to identify promising audiences. A live event organizer may need to understand local taste. And, yes, one may even dream that the makers of elevator systems might someday invest in better musical choices. In that sense, music recommendation is not a narrow application. It is a broad and often underestimated infrastructure for cultural selection and experience.

I confess that I find Spotify’s recommender systems to be among the most useful recommender systems in everyday life. Few systems are so regularly present, and so often genuinely helpful. They can surface forgotten favourites, support exploration, and sometimes offer exactly the right soundtrack for a particular moment. But of course, even good things can be improved. This is precisely why music recommendation remains such an important research topic.

The challenges are substantial. Music catalogues are vast. Feedback is mostly implicit and often noisy. Preferences shift quickly, across hours, days, moods, activities, and life phases. Music is tied to emotion, identity, and context in a way that makes simple relevance optimization look rather naive. On top of that, there are cold-start problems, regional and cultural differences, and the difficulty of evaluating not only accuracy, but also discovery, user experience, and long-term value. The call states these issues very clearly, and that is one reason why I find this special issue especially timely.

The special issue is edited by an excellent team: Christine Bauer, Elena V. Epure, Andrés Ferraro, and Lorenzo Porcaro. This is a strong combination of academic and industrial expertise, and it gives confidence that the issue will attract both methodological depth and practical relevance.

The scope is broad and very well chosen. Topics include evaluation methods, bandits and reinforcement learning, large language models, multimodal and cross-modal recommendation, sequential models, playlist generation, conversational systems, music search and discovery, listener intent and mood modelling, recommender systems for live music and record labels, recommender systems for music creation and generative workflows, and questions of fairness, transparency, explainability, cross-cultural recommendation, and AI-generated music.

The deadlines are also clear. Abstracts are due on May 7, 2026, and full submissions on May 14, 2026. First-round decisions are expected in July 2026, revisions are due in September 2026, and final decisions are planned for December 2026.

For researchers in recommender systems, this is an attractive call. Music remains one of the best domains in which to study what recommendation really means when human taste is unstable, context-dependent, and culturally shaped. It is difficult, messy, and important. Precisely for that reason, it remains one of the most interesting areas in our field.