The button looks harmless: with a flip of a switch labelled “Autoplay,” Spotify is set to “automatically plays similar songs so the music never stops” after the end of your playlist.
But the app that decides what song will play next could be influencing your tastes and the popularity of certain types of music over others. A recent white paper affiliated with the University of Toronto and the Canadian Institute for Advanced Research (CIFAR) examines this very phenomenon. The paper, published by the Schwartz Reisman Institute for Technology and Society (SRI), provides perspectives from Toronto’s music industry.
Machine learning is at the core of how music streaming services listen to the listeners and decide your upcoming queue. Spotify’s engine in particular, according to Hacker Noon, applies a blend of at least three strategies: studying listener behaviour, scraping the web to analyze writing about musical tracks to better categorize them, and listening to raw audio to categorize new songs.
“The design of music recommendation algorithms embeds theories of listening and theories of music, whether intentionally or not,” wrote Dr. Ashton Anderson, professor at the UofT Department of Computer Science and co-author of the paper, to The Strand. But a side effect of this strategy of engineering recommendations is its capture of users in a filter bubble: an isolated pocket whose existence is unknown to many casual listeners.
The narrowing of cultural exposure driven by algorithms is a problem recognized by Canada’s federal government. “Users rely almost exclusively on these algorithms [for content curation] and many voice concerns that only global content from culturally dominant countries is prominent on digital platforms,” according to a 2019 government report highlighted by the CIFAR paper. “It is even more challenging for people from a minority language or from a minority cultural community.”
To strengthen support for musicians willing to experiment to create a diversity of music, Dr. Jeremy Morris, white paper co-author and professor of Media and Cultural Studies at the University of Wisconsin-Madison, said in an interview with The Strand that streaming services can learn from embedding musicologists and social scientists in their engineering teams.
Changes in user interfaces could also help listeners realize when they are in a filter bubble, which could nudge them to explore new genres, noted Dr. Anderson. “If our hypothetical listener is aware of being stuck in a filter bubble, they are off to a good start,” he wrote.
The importance that artificial intelligence plays in music discovery is shaping the ways that artists create their music, with reflections of the academics’ predictions being seen in the experiences of members of the music industry. Artists and labels alike are rewriting their game plans to maximize their chances of success in the highly competitive and volatile game that is the music production industry.
Matt King, a multi-instrumentalist of the band Absolutely Free, notices how, for instance, songs have progressively shortened in length over the years. The strategy likely stems from a desire to maximize the monetary value of songs by increasing the number of plays and streams they receive, he said to The Strand. Very few audiences would listen to a ten-minute song on repeat, but a two-minute and 38-second song such as “Old Town Road” is easier to listen to over and over again.
Music platform technologies not only impact the way songs are created, but also affect how music executives market their artists. As a manager and label owner, Noah Fralick of Huxley Management said to The Strand that his job requires him to ask questions like, “Are there spoken songs on the record that are going to sit in certain playlist spaces? Is this more of a Spotify band or more of an Apple Music band?” From a marketing perspective, such questions are essential, Fralick says, to identify ways to promote the artist that are tailored to the specific platform.
Navigating algorithmic music platforms, managers can find themselves in a difficult situation where they have to weigh the pros and cons of allowing artists to express a unique style and timbre. Distinctness in musical style may be fulfilling for an artist’s personal creative pursuits, but may be buried under the more conventional styles and genres the algorithm is used to promoting, and thus deem the artist less marketable.
Nick O’Byrne, of the artist management group Look Out Kid, comments in an email to The Strand, “All of our artists have a compelling point of difference (language, genre, vocal timbre, lyrical content, instrumentation, etc.) which to their audiences is a strength, but to many mainstream gate-keepers, playlist-editors and algorithms probably make them an unsafe bet.”
Some artists, however, refrain from designating algorithmic music platforms as a hurdle to genuine musical creativity. Musician Jesse Crowe reflected to The Strand that they view musical composition as an opportunity for creative freedom, rather than an endeavour for monetary earnings. Crowe therefore does not follow the statistics of their songs on Apple Music or Spotify, limiting these metrics’ influence on music composition.
But for artists whose livelihood depends on the success of their musical career, the stakes are higher. As a manager, Fralick finds many budding musical artists attached to an illusion of finding success and reaching a massive global audience through streaming platforms. He likens the chances of making it big on such a platform—stream count-wise—to winning a lottery. And similar to a lottery, the wins are broadcast in clear daylight for the world to see, while the countlessly larger number of losses get pushed aside.
Fralick adds that artists get a momentary high when their songs make it into highly streamed playlists, or get “playlisted”—though the boost in listeners is often short-lived, as the playlist regularly can change. King, who creates music for personal and spiritual purposes, states that for artists striving to find considerable monetary success on a streaming platform, it is as difficult as “finding a diamond in the rough.”
So how should artists navigate the murky waters of algorithm-based music distribution? King suggests that one promising approach is to forge real connections outside of the algorithmic world. Crowe corroborates this method, citing that they met their first collaborators at a birthday party. It seems that just as is the case in most other fields, the sustainability of a musical career depends more on steady real-world progress than on bursts of online hits.
The opaqueness of the workings of artificial intelligence itself means that its full ramifications on the music industry have yet to be seen. In the meantime, King encourages artists to “just keep doing what you do. And if it’s good, maybe somebody will find you, maybe not.” Often acting in unexpected and bizarre ways, streaming platforms offer no fool-proof formula for success that artists can follow. Recognizing this uncertainty, King asks, “Who knows where the algorithm is going to pull people to?”
For music listeners, the traditional methods of music discovery offer exciting ways to escape the filter bubble, reflected Dr. Morris. “When I lived in Canada and was in Toronto and Montreal, I felt really a part of those local music scenes,” he said. As Toronto opens back up, small local venues could be the place where you can find your next favourite band.
