Strategies for mitigating artist gender bias in music recommendation: a simulation study

Christine Bauer*, Andrés Ferraro

*Corresponding author for this work

Research output: Chapter in Book/Report/Conference proceeding/Legal commentaryConference contributionpeer-review

Abstract

Recommender systems are prone to various biases. Hence, bias mitigation approaches are needed to counteract those. In the music sector, gender imbalance is a particular topical subject. Earlier work has shown that the gender imbalance in the sector translates to the output of music recommender systems. Several works emphasize that items representing women should be given more exposure in music recommendations. In this work, we present an exploratory analysis of several bias mitigation strategies. Using a simulation approach, we explore the effects of different pre- and post-processing strategies for bias mitigation. We provide an in-depth analysis using state-of-the-art performance measures and metrics concerning gender fairness. The results indicate that the different strategies can help to mitigate gender bias in the long term in particular ways: Some strategies' render improvement in exposure of women in the top ranks; other approaches help recommending more variety of items representing women.
Original languageEnglish
Title of host publicationMuRS: Music Recommender Systems Workshop
PublisherZenodo
Number of pages5
DOIs
Publication statusPublished - Sept 2023
EventMuRS: Music Recommender Systems Workshop - Singapore, Singapore
Duration: 19 Sept 202319 Sept 2023
Conference number: 1
https://sites.google.com/view/murs

Workshop

WorkshopMuRS: Music Recommender Systems Workshop
Abbreviated titleMuRS
Country/TerritorySingapore
CitySingapore
Period19/09/2319/09/23
Internet address

Keywords

  • recommender systems
  • artists
  • music
  • music recommender systems
  • bias
  • bias mitigation
  • fairness
  • gender balance

Fields of Science and Technology Classification 2012

  • 102 Computer Sciences

Cite this