Tracing Desired, Exposed and Adopted Algorithmic Novelty on Music Streaming Platforms

Almost a decade after the Filter Bubble’s instantiation, concerns that algorithmic guidance may limit diversity remain as widespreadas ever. Nonetheless, supporting empirical evidence has, over the years, weakened. Across some of the most widely adopted search engines and social media platforms, several works have found that users’ own selections of algorithmic recommendations may principally be responsible for constraining the diversity of media which they end up consuming. Phrased differently, users may be in what literature has coined, “self-imposed filter bubbles”. On music streaming platforms, similar concerns have frequently been raised over the role which algorithmic guidance plays in influencing users’ consumption behaviours particularly in terms of novelty and diversity. Nonetheless, existing empirical literature geared at appraising these dynamics generally leverage solely consumption logs and thus fail to disentangle the impact of user choices from algorithmic decisions. Drawing on both behavioural and survey data from users of the French music streaming platform Deezer, we examine the relationship between the levels of algorithmic novelty users desire, the levels exposed to them and the levels ultimately selected to be consumed. In turn, our work demonstrates that despite commonly expressing a desire for balanced novelty, Deezer users oppose algorithmic novelty, selecting less novelty to consume than what is exposed to them algorithmically. At the same time, our work underscores the complexities of designing platform strategies geared at nudging users towards diversity.