Social music discovery: an ethical recommendation system based on friend's preferred songs

Furini, Marco

DOI: https://doi.org/10.1007/s11042-024-19505-0

IF: 2.577

2024-06-12

Multimedia Tools and Applications

Abstract:Music recommendation systems have become ubiquitous in today's world, but they raise ethical concerns related to bias, discrimination, and lack of transparency. To address these issues, we propose a recommendation system that combines content-based and collaborative filtering approaches within three different recommendation algorithms. These algorithms create playlists that mimic the user's listening habits while identifying similar tracks within the listening histories of the user's friends. To evaluate the effectiveness of our system, we asked ten participants to rate a total of ninety playlists. The results showed high satisfaction among participants with the playlists generated by two of the proposed recommendation algorithms. Specifically, participants who preferred to stay within their musical comfort zone appreciated one specific recommendation algorithm, while those who were willing to explore new music tended appreciated the other recommendation algorithm. In summary, by leveraging the user's social connections, our proposed system provides a more transparent and ethical approach to music recommendations. It provides a personalized and enjoyable music discovery experience that considers the nuances of individual musical taste and preferences. These findings suggest the potential impact of our proposal in addressing ethical concerns and enhancing user satisfaction in music recommendation services.

computer science, information systems, theory & methods,engineering, electrical & electronic, software engineering

What problem does this paper attempt to address?

The paper attempts to address ethical issues in music recommendation systems, particularly concerns related to bias, discrimination, and transparency. Specifically, the authors propose a recommendation system that combines content-based and collaborative filtering methods, utilizing users' social relationships to generate personalized playlists, thereby introducing greater transparency and fairness into the recommendation algorithm. The system aims to create playlists that mimic users' listening habits by identifying similar tracks in the listening history of their friends, and the effectiveness of this recommendation mechanism is validated through experiments. The study finds that different recommendation algorithms are welcomed to varying degrees based on users' music preferences and willingness to explore new music. Overall, the system provides a more transparent and ethical way of music recommendation by leveraging users' social connections, enhancing user experience and satisfaction.