Whenever I reach the end of one of my playlists, Spotify goes ahead and generates a queue of recommended suggestions. K cool. However, it's gotten to the point that I can predict 95% of the songs that it's going to play.
As a feature that is supposed to help me discover new music (or at least, not actively infuriate me) all it is doing is overplaying and ruining songs. The algorithm will pick up a song that I have just listened to, and play it again, and again. Maybe every 25th tune will be one that I haven't heard before (if I'm lucky) or at least something that I haven't heard for a while. As a success ratio for predicting like songs, that's atrocious.
It's mostly incredibly frustrating because I know how much data is available to your platform, and it's a bit of an insult that your ML team cannot even get the basics of "don't play the same song three times an hour" down. There are people who would sell their firstborn to feed their models with your data, so it's like, seriously can you not do better?
So obviously you're not going to build a better model just because I'm annoyed (and yeah I'm sure you're working on it, anyway). But if there could at least be a way of "banning" songs, that would be a good intermediary step. Some kind of denylist to take overplayed tunes out of the pool. Even a way of setting a filter for possible suggestions, eg: "as a user, if I have listened to a song x times in last month, don't shove it in my face for the nth time." Something like that.
There used to be the thumbs-up, thumbs-down buttons (which I'm gonna bet were primarily aesthetic) so could we not have something like that back? A "click this and we won't suggest this song again" button. Users who use the platform heavily would not mind providing feedback if it genuinely resulted in an improved listening experience. I am such a user. I would not mind.
Thanks. And sorry ML team, I'm sure you guys are great. I get that recommendations tend towards silos. But with the resources at your disposal, as the biggest music streaming platform in the world, I gotta believe that you can do better.