User-Driven Fairness in Music Recommendations: Effects on Experience

Abstract

This study investigates how user-driven customization of fairness and diversity affects satisfaction in music recommender systems. We developed a prototype allowing listeners to adjust four fairness dimensions: popularity, artist gender, nationality, and genre diversity. In 42 sessions with Dutch participants, interactive controls substantially improved perceived fairness, control, and added value. Genre diversity was most influential, while nationality was least engaged, with gender and popularity falling in between. Findings highlight that parametric design-not algorithmic complexity-drives improved user experience. We show that transparent, customizable fairness levers can make recommendation systems both fairer and more engaging.