Breaking feedback loops in recommender systems: an example on mitigating artist gender bias in the music domain

Abstract

Algorithms have seamlessly integrated into the music industry, with music recommender systems facilitating navigation through vast catalogs of music tracks. These systems suggest similar artists or recommend the next track for us to listen to. An ideal music recommender system should recommend the ‘right music to the right person at the right moment.’ However, what happens if it falls short of being ideal? In this presentation, I delve into the perspective of artists. Among other aspects, I will present research findings on artist gender bias in music recommendations and provide strategies for mitigation.