ALGORITHMIC CLASSIFICATION OF MUSICAL FILES: User Perceptions and Influence on Listening Practices
Media and Music, Musical Genre, Actor-Network Theory, Algorithmic Culture, Recommendation Systems, Human-Computer Interaction
This research employs Actor-Network Theory (ANT) to analyze music streaming as a complex sociotechnical system, where interactions between agents, such as users and algorithms, are studied as a feedback relationship. The streaming platform is seen as a controller of online flows, addressing the user-algorithm
relationship and the influence of technical objects on subjectivity and musical enjoyment. The symbiotic relationship between the algorithm and the user is highlighted, with the latter playing a crucial role in the continuous development of systems by allowing themselves to be monitored. The work explores the concept of musical genre and the classification practices of musical genres by artificial intelligence algorithms on streaming platforms. The fluidity and intersections of genres are emphasized, with ANT aiding in musical analysis by understanding these multifaceted relationships. Thus, the fluidity of the musical genre concept is discussed, merging streaming data with the perspectives of Latour, Piekut, and Drott to propose a flexible analysis of genres. Additionally, the work (1) explores how user interactions on Spotify shape the musical experience, emphasizing personalized recommendations and filtering techniques. (2) It addresses the evolution of musical classification practices on Spotify and the analysis of musical properties by audio processing algorithms. (3) It examines extreme personalization and its impact on musical exploration, highlighting how the user-algorithm interaction can limit musical diversity. The text also (4) presents impressions from interviewed users regarding the ways musical classification on the streaming platform's interface, and how these classifications are negotiated by users during their interactions with the platform.