For the past few days, I’ve been trying to use a music streaming service—specifically Beats Music—as my primary way of listening to music. I have a number of thoughts that I’m still working through, but there is a design pattern I’ve noticed, common to all the major streaming services, that deserves some scrutiny. On Beats music, it looks like this:
Other services use the Siskel and Ebert “Thumbs Up/Thumbs Down” approach, but the idea is the same. It asks: “Do you like this song, or do you dislike it?” Your selection is tied to whatever algorithmic method the service uses to determine what else it should play for you, though it’s not terribly explicit how it works. I think it’s safe to assume that a Heart or Thumbs Up means “Give me more of this,” and the X’d Out Heart or Thumbs Down is “Give me less of this.” Opting not to choose, I imagine, is interpreted as apathy.
The problem is that I don’t relate to a music on the binary level (or trinary, if you want to count the “meh” option of not tapping either) that these services use. My feelings about music run along a fairly wide scale that, at the extremes, could be described as “I want to hear this song over and over again until I drive everyone around me insane” to “If I hear this song again, I will punch someone.” This is probably too wide, and personal, of a gamut of values to expect a streaming music to implement, but that doesn’t mean there can’t be more nuance involved.
It’s possible there already is. If the folks handling these algorithms are as smart as they think they are, they’ll probably assume that a Thumbs Down, followed by skipping to the next song would weigh heavier against a track than a simple Thumbs Down, or that a Heart followed by a skip would indicate “I like this song, but I don’t want to hear it now… maybe play this artist less often.” These algorithms are so opaque, however, that there’s no way of knowing for sure. That’s what really throws me. How many times do I need to tap the X’d Out Heart and then skip the song in Beats for it to know that I really do not want to ever have to hear INXS again, ever?  It would be better if I could be just a little more explicit in how I feel about a song or an artist, so that I don’t need to rely 100% on the algorithm’s learning process.
After all, if algorithms are going to play such a massive role in what media we’re exposed to, it benefits us to have some insight into how they decide, even if it’s to correct the inevitable mistakes it’ll make during the learning process—and, occasionally, after. I want to see what my streaming service thinks about me, something like Google’s Demographics page meets Last.fm. I’d like to see things broken down by genre, by artists, possibly even by song, if I’ve made any specific decisions on any. That’s one of the biggest issues I have with streaming music—i’m wary of giving up control of the music I listen to purely to some black box of an algorithm. Probably because so many of them are both opaque and inaccurate all at once.
One point in Beats favor is that it’s onboarding process includes a step where you explicitly exclude some genres and artists, but it’s hardly comprehensive. I’m just glad it gave me a chance to banish The Smiths before I heard note one of Morrissey’s atonal warbling. ↩