One of my favorite things about music streaming services is the ability for algorithms to sort through the tens of millions of songs on the platform and pull out the ones I am most likely to enjoy. These algorithms aren't perfect, but they're definitely useful high-level filters. That being said, I noticed something peculiar on both Apple Music and Spotify today, and once I started digging into it, it pointed toward something quite revealing about the direction these algorithms could push our listening habits.
I, like many angsty bedroom guitarists, am a big fan of the indie rock group Pavement. Songs like "Cut Your Hair," "Here," "Spit on a Stranger," and "Summer Babe (Winter Version)" are some of the sharpest alternative songs ever written and only seem to be growing richer with age. But strangely enough, despite all of these songs being uber-popular bonafide classics, none of them hold the spot of "top Pavement song" on Apple Music or Spotify. No, this spot belongs to the obscure b-side "Harness Your Hopes," a song which almost no fan would rank as the best Pavement song and many casual fans don't even know exists. In fact, "Harness Your Hopes" actually holds the top two spots in Pavement's top songs on Apple Music: a live version for the BBC at #1, and the studio version at #2.
I always assumed this song must have been in a famous movie soundtrack or something and I was just out of the pop-culture loop, so I never questioned it. But today I decided to question it, and all it took was a quick Google of the song title to get my answer.
The robot DJs are up to their old tricks again.
The first thing you get when you search "Harness Your Hopes" in Google (except for the Genius lyrics) is a Stereogum article titled "Why Is The Obscure B-Side “Harness Your Hopes” Pavement’s Top Song On Spotify? It’s Complicated." The piece explains that "Harness Your Hopes" was never high in Pavement's song rankings on streaming services until 2017 when Spotify launched its "Autoplay" feature. This feature allows a listener to forget about building out a queue manually and instead let Spotify algorithms create a queue of songs that are sonically similar to the one you're currently listening to. In essence, it's a robotic DJ that always keeps the vibe going. For one reason or another, this Autoplay algorithm was disproportionately recognizing "Harness Your Hopes" as a song that sounded similar to a lot of other songs, leading to it getting played an awful lot and ending up in a lot of users' playlists and libraries despite never being a single or even a popular deep cut for the band.
For as long as music has existed, the artist or music industry has always had a say in telling us what "good music" is. The clearest example of this is in labels and artists earmarking some songs as "singles" as if to signal, "These are the songs I am betting my entire artistic reputation on." This Pavement incident proves this taste-setting concept could be on its way toward extinction, as the music industry clearly doesn't have nearly the leverage to dictate what songs get the most attention on a release now. This sounds great on the surface in a very "democratize music listening" way, but this does bring with it a pretty significant downside...
The downside is this: "Harness Your Hopes" is not Pavement's best song. It's just not. And that's not me annoyingly inserting my opinion into a rather objective argument: the Pavement classics I mentioned above like "Summer Babe (Winter Version)" and "Spit on a Stranger" had two decades to rise above the pack and distinguish themselves as the cream of the crop. An obscurity like "Harness Your Hopes" is the "top Pavement song" sheerly based on exposure, and it's hard to argue that anything is more important than exposure in predicting a song's popularity. You'll grow to like almost any half-decent song if you listen to it enough, but that's not a great way to build-out your music taste. And a song like "Harness Your Hopes" getting this level of exposure wouldn't be that problematic if the reason it was getting this exposure wasn't just because it "sounds like a lot of other songs." If the algorithms are optimizing for this goal, won't that inevitably lead to a sort of "converge to the median" in recommended music?
I'm all for taking power out of the hands of labels and music industry executives and giving it back to the artists and the listeners. But the truth of the matter is labels were and still are really good at picking singles. Labels know what will most likely connect with an audience, and they do what they can to make sure that song gets to your ears first, even if occasionally an album track will beat the odds and become the clear fan favorite (looking at you, "All Too Well"). Clearly robots aren't quite getting the job done, as designing artist profiles so two versions of an obscure b-side become that artist's two highest-ranked songs is doing a disservice to both the misrepresented artist and the underwhelmed listener whose first impression of the artist could've been stronger. I would love to see Spotify and Apple Music continue working to return a bit of that control from the robots back to the artists, who can then have a bigger role in curating which of their songs are prioritized in algorithmic recommendations and their artist profile song rankings.
Comments