Algorithms tailoring music recommendations for listeners around the world are more likely to feature music created by male artists — preventing content by women artists from reaching a wide audience, a new study has found.
Published by the Association for Computing Machinery, the study says on average, male artists feature on top of recommendation by streaming platforms, whereas women come in at the sixth or seventh position. In addition, the proportion of songs by women on these lists is also much lesserthan music by men.
The music industry is already known for its gender disparity — female musicians, producers, and label executives continue to be underrepresented; all while the gender pay gap widens. Researchers note the algorithm mirrors the existing bias in the dataset of the study, in which only 25% of artists were women — echoing the underrepresentation of women artists in the industry. As such, gender-biased recommendations by algorithms can further entrench the discrimination, especially with more and more users switching to music streaming platforms.
Algorithmic biases not only minimize the exposure female musicians receive in the short term, but have a wide-ranging impact. Recommendation systems are self-evolving — algorithms learn from the choices people make and are designed to offer future suggestions based on what people choose. If users keep selecting male artists who top these lists, the recommendation algorithm will show them more of this category. In the long run, algorithms can bury content by women deeper down the lists, creating a vicious cycle of what some call a ‘feedback loop’.
Related on The Swaddle:
Men, Women See Different Job Listings on Facebook Due to Algorithm Bias, Research Shows
Interestingly, since music listeners appear to rely heavily on recommendations, the same feedback loop can help fix the bias too — by simply recommending more songs by women. In fact, the researchers tested whether re-ranking the gender-biased recommendations can affect users’ listening behavior, and the results came out positive. “Simulating the feedback loop shows that gender can be balanced in the long term by gradually increasing the exposure of female artists in the recommendations,” the study noted.
“…the population of the world is 50% women. So it would be ridiculous if the system wouldn’t recommend them,” one of the many artists interviewed for the study, who acknowledged the prevalent gender bias in music recommendations, said.
An endeavor that ensures gender fairness can however be met with strong opposition from listeners — they may resist such a change that offers ‘positive disparate treatment,’ or positive discrimination that helps uplift minority sections, in favor of female artists. To counter this, the study emphasizes the need to roll out positive anti-discrimination policies, but “only gradually until gender balance is reached to avoid reactance,” since users are key to achieving the goal of equitable gender representation.
