A recent study by the Music Technology research group with the University of Utrecht indications shows that a widely used recommendation algorithm is more likely to choose music by male artists than female artists. Andrés Ferraro and Xavier Serra, researchers of the Music Technology research group (MTG) with Christine Bauer, of the University of Utrecht (Netherlands), published a paper on gender balance in music recommendation systems in which they ask themselves how the system should work to avoid gender bias.
Initially, the work aimed to understand the fairness of music platforms available online from the artists’ point of view. They identified that gender justice was one of the concerns.
The authors tested a commonly used music recommendation algorithm based on collaborative filtering and analyzed the results of two datasets. In both cases, they found that the algorithm reproduces the existing bias in the dataset, where 25% of artists are women. The algorithm generates a ranking with artists to recommend to the user. On average the first recommendation of a woman artist comes in 6th or 7th position, while a man is in the first position.
The situation only gets worse when taking into account that as users listen to recommended songs, the algorithm learns, creates a feedback loop—and they get more of the same.
The authors of the study propose a new approach to allow greater exposure of female artists and that would consist of reordering the recommendation, which would move a specified number of positions downwards in order to solve the existing gender bias.
In a simulation, the authors studied how classified recommendations would affect user behavior in the long run. The reclassified algorithm tends to change their behavior. The new algorithm, based on machine learning, would consolidate this change in behavior.
Andrés Ferraro, Xavier Serra, Christine Bauer (2021), “Break the Loop: Gender Imbalance in Music Recommenders”, a Scholer F, Thomas P (eds.). CHIIR ’21: Proceedings of the 2021 Conference on Human Information Interaction and Retrieval. 1 ed. New York: Association for Computing Machinery; 2021. p. 249-254. https://doi.org/10.1145/3406522.3446033
Original Source: Eureka Alert