On Pitchfork, Jeremy D. Larson writes about “The Woes of Being Addicted to Streaming” and makes an interesting observation about how yet another algorithm—that of music recommendations—has is influencing global taste:
The more time I spend on Spotify, the more it pushes me away from the outer edges of the platform and toward the mushy middle. This is where everyone is serviced the same songs simply because that is what’s popular.
The Spotify algorithm will, of course, recommend you music based on what you’re already listening to. But in order to make that recommendation, it uses data on which songs are the most popular. But what if what’s popular on Spotify isn’t what’s normally popular? What if the recommendation has artificially created a popularity by recommending a song so often that it becomes a hit?
He goes on:
Four years ago, while the app’s algorithmic autoplay feature was on, I was served the Pavement song “Harness Your Hopes,” a wordy and melodic—and by all accounts obscure—B-side from the beloved indie band. As of this writing, the song has over 72 million streams, more than twice as much as their actual college rock hit from the ’90s, “Cut Your Hair,” the one Pavement song your average Gen X’er might actually recognize. How did this happen? In 2020, Stereogum investigated the mystery but came up empty-handed from a technological perspective, though the answer seems obvious to me: Whereas many Pavement songs are oblique, rangy, and noisy, “Harness Your Hopes” is among the most pleasant and inoffensive songs in the band’s catalog. It is now, in the altered reality of Spotify, the quintessential Pavement song.
The algorithmic recommendation can create an alternate reality of what’s popular, then proceed to serve it to its audience until it actually becomes true. Obviously, the people listening must still like what they’re hearing, but it’s nevertheless curious to think of these alternative universes of popularity, especially when there are several dominant platforms.
The whole “Harness Your Hopes” situation is in part a result of what’s called “cumulative advantage.” It’s the idea that if something—a song, a person, an idea—happens to be slightly more popular than something else at just the right point, it will tend to become more popular still. (On the other hand, something that does not catch on will usually recede in popularity, regardless of quality.) This is the metric of how most social recommendation algorithms work—on Facebook, the more “likes” an article has, the better odds a user will read it. But when this is applied to what songs are sent to which people, Spotify can engineer its own market of popularity as well as what song defines a band.