Does Spotify’s algorithm pose an existential threat to our music taste?
1s and 0s more like why does everyone listen to the same stuff now
By Laura Weinstein
I grew up fingering my way through my father’s record collection. From The Doors to Gordon Lightfoot to The Police and Radiohead, I knew there were limits to what I could find. As the years went by, my father invested an appalling amount of money into stereo equipment and a vast array of archival records to parse through in the name of audio quality and physical ownership of media. Seeing his passion for music manifest in a collection of neatly sorted records and shiny new stereo equipment was awe-inspiring. But when I got my hands on my first laptop in 2013, an infinite supply of digital media consumed me. From my twelve-year-old perspective, crate digging became obsolete. Why would I spend my time sorting through records when, with the click of a few keys, I could find the track I heard on a GEICO commercial, listen to the entirety of the new The 1975 album for free, and have other similar albums suggested to me too? It became a no-brainer to turn to the internet for all my music listening needs.
In the past ten or so years, my audiophile father got with the times and began playing music on his speakers through a high-fidelity amplifier using the streaming service Tidal. But with a change in recorded medium, his listening became formulaic. The more he used Tidal, the more I realized how similar all alternative rock suggestions are– he was being recommended the same songs as I was. Our taste began to merge in a way it never had before he used streaming services. As he stepped out of record stores and into the digital ether, I began to see how much our taste in music was predesignated. The second my shoegaze playlist ends, the same songs every other 21-year-old with a thing for dissonant chords begins. And it begs the question– does Spotify’s algorithm pose an existential threat to our music taste?
Spotify has not been the most transparent about how its algorithm works. What we do know is that it processes a large amount of user data to recommend songs. Additionally, it categorizes songs based on danceability, emotional tone, mode, chord progression, complexity of melody, and many other characteristics. It also uses natural language processing to recommend songs based on narrative structure, themes, and underlying sentiment of lyrics and it does a hell of a good job of doing it. Spotify has gone as far as to create playlists for each user that are updated to match the weather, location, and time of day. As much as I am ashamed to admit it, more than a quarter of the songs I listen to have been recommended to me by this algorithm.
But where the algorithm often goes wrong is with its collaborative filtering feature that has a seemingly random bias towards particular songs. An example is Pavement’s “Harness Your Hopes.” In a 2020 Stereogum article, Nate Rogers was one of the first to speak up about this phenomenon. What was originally an overlooked b-side track off of Pavement’s 1997 record Brighten the Corners is now their most played song by tens of millions of streams. Currently at 113 million streams on Spotify, “data alchemist” Glenn McDonald from the company says the ‘Auto-play’ feature is to blame, or thank, for this. After the feature launched in 2017, “Harness Your Hopes” rapidly started climbing the charts. And in 2020, the song began trending on TikTok, launching the deep track out of obscurity and into the playlists of teens and young adults everywhere.
The flip side of this phenomenon can be illustrated by indie folk/rock artist Alex G after his unreleased song “Treehouse” from his pre-Domino Records days went viral on TikTok. The track wasn’t even on Spotify when TikTok users began using the song in 30-second edits of their quarantined lives in 2020. After years of fans pestering the artist to put this one-off track on Spotify, he caved in and probably made a decent buck off of it. As of December 15th, 2023, at 134 million streams, “Treehouse” is G’s second-most-played song on Spotify.
In addition to the Spotify-to-TikTok and TikTok-to-Spotify feedback loop, recommendation features have a bias against new artists with little user data to work off of and artists of marginalized identities. As the algorithm is based on user data, it is not immune to our own cultural biases. As Spotify learns from our listening habits, it perpetuates the cycle of undiversified listening.
I have no issue with my favorite lesser-known artists going viral, but I do take issue with the fact that I’m allowing my musical taste to be dictated by statistical trends. With our overreliance on math equations, we are ignoring a large selection of music. So what is the solution? Do we abandon using Spotify’s recommendation features altogether?
Overcoming Spotify’s algorithmic biases lies in the hands of product managers and software engineers, but on a personal level, exploring music more holistically is one way to combat this cultural crisis. Paying attention to the music I hear in public is one major way I discover songs, as it probably is for many others. Whether I’m in a cafe or a restaurant or am overhearing someone’s car radio, it is surprising how many gems are hidden in plain sight.
Of course, sitting through entire albums rather than pressing shuffle on musicians’ Spotify pages is another way to beat the algorithm. Regularly exploring different genres of music, even the stuff you swore you hated is a healthy practice as well. Crate-digging, although tedious, can be very rewarding; owning physical media also compensates artists more than streaming does, another incentive to lay off the Spotify– but that’s for another article.