Music recommendation algorithms show gender bias

Although the problem of gender discrimination already flourishes in the music industry, the study by researchers at Pompeu Fabra University (UPF), Barcelona, and the University of Utrecht (UU), Netherlands, found that music recommendation algorithms are increasing the gender gap.

Andrés Ferraro and Xavier Serra, the MTG researchers at UPF, along with Christine Bauer of UU, recently 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 by Ferraro, Serra, and Bauer aimed to understand the fairness of music platforms available online from the artists’ point of view. In interviews conducted with music artists, they identified that gender justice was one of their primary concerns.

The team tested a commonly used music recommendation algorithm based on collaborative filtering and analysed the results of two datasets. In both cases, they found the algorithm reproduces the existing bias in the dataset, in which only 25 per cent of artists are women.

In addition, the algorithm generates a ranking of artists to recommend to the user. Here, the researchers found that on average the first recommendation of a woman artist comes in 6th or 7th position, while that of a man artist is in first position. “The bias in exposure comes from the way it generated recommendations,” Ferraro explained, meaning that women have less exposure based on the recommendations of the system.

The researchers said the situation worsens when considering that the algorithm learns as users listen to recommended songs. This in turn creates a feedback loop. But with the help of the reordered algorithm, users change their behaviour so that they listen to more female artists.

The researchers have proposed an alternative approach that would allow greater exposure of female artists and would comprise reordering the recommendation, which would move a specified number of positions downwards in order to solve the existing gender bias.

In a simulation, the team studied how classified recommendations would affect user behaviour in the long run. The results showed that, with the help of the reclassified algorithm, users would change their behaviour and thus listen to more female artists than with other music recommendation algorithms, and the new algorithm, based on machine learning, would merge this change in behaviour.