Spotify is trialing a new ability for users to turn off their personalized recommendations, according to beta code unearthed by MacRumors contributor Aaron Perris.
Depending on how it works, the new option could conceivably allow users to prevent songs played on their account on behalf of other listeners from affecting their personalized recommendations. Either way, it is likely to be welcomed by Spotify users who aren't entirely satisfied with the success rate of the service's algorithms in matching new songs to their listening tastes.
Spotify's recommendation system combines a number of machine learning and AI-driven methods. The service's algorithm analyzes listening history, favorite genres, and the time of day that users listen to music, and bases its personal recommendations on these parameters. This process is combined with content-based filtering, which analyzes the audio of a song that a user interacts with, and takes into account characteristics like time signature, key, and tempo to recommend new songs with similar characteristics.
Spotify's recommendations system also uses something called collaborative filtering. For example, if two users have similar music tastes, Spotify might recommend songs one user has listened to but the other hasn't. However, this can end up with Spotify recommending more popular items more often than lesser known songs, resulting in an "echo chamber" effect, and for new users with no listening history it is next to useless.
It's not clear when the new option to turn off personal recommendations will appear for Spotify users, or whether the trial is region-specific. Spotify simply says it is "evaluating the ability," and thanks users for trying it out to help make Spotify better.
Apple is introducing a similar function for Apple Music subscribers in iOS 17.2, in the form of an additional Focus filter that can automatically turns off a user's Apple Music listening history. Apple's substantial software update is expected to land on iPhones in December.
