Proceedings of the 37 th International Conference on Machine Learning
To evaluate if the recommendations are fair, we have to consider how all the stakeholders are affected. In this work, we focus on the artists in the music domain. We analyze the recommendations made with Collaborative Filtering from the artists¿ side to understand how the recommender system can affect the artists¿ reach and exposure. To this end, we group the artists using different aspects: location, gender, period, and type (e.g., solo, band, orchestra) and study the effect of the recommendations on these groups, comparing their distribution in recommendations, created by the system, with the previous activity of the listeners.
Ferraro, A.; Jeon, JH; Kim, B.; Serra, X.; Bogdanov, D.. Artist biases in collaborative filtering for music recommendation. In: ICML. Proceedings of the 37 th International Conference on Machine Learning. 1 ed. 2020.