Spotify is testing a new feature that gives users direct control over their recommendations. Initially available for Premium subscribers in New Zealand, the tool allows adjusting the Taste Profile by indicating preferences for genres, artists, or podcasts. You can request more or less of a specific content. Additionally, filters for podcasts are added. The goal is to complement the algorithm with manual choice.
Direct feedback interface and algorithmic adjustment 🎛️
The feature operates as an explicit feedback layer over the recommendation system. Adjustments a user makes to their profile would act as high-weight signals for the algorithm, readjusting the priority of certain tags or attributes in content selection. It does not replace the base machine learning, but introduces a manual correction parameter. Podcast filters (unheard, in progress, video) simplify management from the client side.
Goodbye to the algorithm's musical scares 😅
Finally, we can tell the system that, although we listened to a reggaeton song at a party three years ago, we don't want it to fill our Daily Mix with that. It's the equivalent of telling a friend with bad taste to stop suggesting things to you. The question is whether, after years of training the algorithm, now we will have to train ourselves to correct it. Extra work, but it might prevent the running playlist from starting with ballads.