Perceptions of Fairness and its Impact on User Choices in Music Recommendation

Abstract

Music recommender systems (MRS) have significantly changed how listeners discover and interact with music. However, these systems often prioritize mainstream content and hence marginalize niche music, as well as users with specific tastes or backgrounds. This imbalance stems from demographic biases as well as algorithmic design, and it may reinforce filter bubbles and limit exposure to diverse music. While previous research mostly addressed fairness in music recommendation from a system-centric perspective, little attention has been given to user perceptions of fairness and how these perceptions affect decision-making. In this paper, we aim to explore participants' initial preferences, how additional information influenced their choices, and the factors motivating changes in decision-making. To address these questions, a mixed-methods user study was conducted with 28 participants, combining quantitative song ranking and categorization tasks with qualitative reflections on fairness categories. The results show that users are generally reluctant to revise their initial choices, even when given more positive or negative information, showing the influence of the commitment principle. By viewing fairness as a user-centered experience rather than merely a system attribute, this study offers new insights for designing recommender systems that better promote diversity and fairness.