April 03, 2024 • By The UCI Paul Merage School of Business
Everyone’s musical tastes are unique and varied, but what prompts us to incorporate new musical styles into our playlists? The answer may simply be about disruptions to our daily routine.
A recent article by Assistant Professor of Teaching Organization and Management Noah Askin of the UCI Paul Merage School of Business, together with James A. Evans of the University of Chicago and Khwan Kim of INSEAD, was inspired by Askin’s lifelong curiosity about the power of music. The initial idea for “Disrupted Routines Anticipate Musical Exploration,” published in The Proceedings of the National Academy of Sciences (PNAS), originated five years ago during a discussion between Askin and Evans at a conference.
“Figuring out why we like the things we like is kind of my holy grail,” says Askin. “But it’s very hard to come up with a grand theory of why you like the things you like or to develop a universal theory.”
Understanding How Tastes Develop Over Time
“I was most interested in determining patterns of consumption,” says Askin. “I was interested in trying to figure out how people develop their listening habits or change what they listen to. I knew figuring out the initial stages of what people like would be very hard, if not impossible, but I thought we could at least contribute something to the understanding of how tastes develop and evolve—and what those changes might be linked to.” That’s what the researchers were most curious about: Do tastes shift toward or away from what’s popular globally?
Using available data from Deezer—one of the largest global streaming services—Askin, Evans, and Kim wanted to see what happens when people travel outside their established environment. “We got subsets of listening data from nine countries and listening histories from 44,000 listeners over three years. That’s hundreds of millions of streams,” says Askin. “Very early on, we brought on Khwan, who’s amazing with computational methods. With his help, we poked around the data to find out what happens when people move around.”
Establishing Geographic Taste Profiles
What the authors eventually developed was a taste footprint for every individual the data represented. “We had to figure out what everyone’s listening patterns looked like,” says Askin. “That taste footprint, or fingerprint, includes all of the songs they listened to. Next, we used a methodology called Song2Vec (S2V), which takes songs in the order they’re listened to in the same way generative AI uses large language models to measure a pattern.” This involves taking a particular word, or piece of a word, and observing the data to guess what word is likely to precede it and what word is likely to follow it. “It’s a contextually based understanding of language.”
They did something similar with a playlist. Given a particular song, could the researchers extrapolate what other songs people listen to? In this way, they could create a listening “fingerprint” for a certain song. This fingerprint is, roughly speaking, an estimation of the likelihood of what’s happening around a given song.
Using this methodology, Askin and his colleagues developed a sense of where a song is likely to be positioned and what other songs are most similar to it based on those patterns. “Once we have that,” Askin says, “we can build up an individual’s taste footprint and determine what their consumption patterns look like. Once we’ve developed that footprint for individuals in a city, we can create an average listening profile for that city. Once we have cities, we can aggregate up to countries.”
How Routines Impact Listening Habits
After Askin and his fellow researchers developed these taste footprints, they were able to see results more clearly. “We could look at how far someone traveled on a particular day,” he says. “If you’re usually spending time in Southern California, but we know you listened to music in London, we can measure how far you traveled and see what you listened to while you were there and how different it was from your usual taste footprint. Was it closer to the general listening patterns of the United Kingdom or the United States?” They were able to gain insight into how much that changed from usual listening patterns in the last six months.
Askin and his fellow researchers found that our cultural consumption patterns—what music we listen to, and by extension what shows we watch and what art we consume—tend to change as a function of our routines. “In other words, the more disrupted our day-to-day activities are, the more likely we are to explore different culture,” Askin says. “As individuals travel away from their homes, and especially as they travel farther and more internationally, they’re more likely to listen to new sounds and expand their musical tastes.”
Disruption in routines can also take place when we stop our normal activities, as many did during the COVID lockdowns. “Similar exploration of musical tastes occurred during the pandemic,” Askin says. “What’s interesting is you might think people just had more time to listen to new music, but that’s not the whole story. There wasn’t that much more listening time going on, at least not in the large sample we had access to.” The driving factor for this increase in musical exploration was a result of the disruption to routines, he says. “As soon as we break from regular activities, we venture out a bit more.”
How Geography Factors Into Tastes
“As people travel farther away from home, their level of musical exploration increases, and the farther away the listening patterns of the city you visit are from your home city, the more likely you are to adapt to those patterns,” says Askin.
For example, he says, a similar distance separates London from Johannesburg, South Africa and Rio de Janeiro, Brazil. A great deal of similarities can be found between the average listening patterns in London and those in Johannesburg, but the average listening patterns between London and Rio are quite different.
“A London traveler who goes to Rio is much more likely to adapt to the tastes and the general consumption patterns of Rio. Their tastes will move further because their taste distance is further than if they travel to Johannesburg. The disparity between the two locations is not about distance; it’s about the disparity of the listening profiles between those two locations,” says Askin. “If they’re not as pronounced, then your travels won’t really change much, but if you travel somewhere where it’s radically different, then it influences you.”
Another fascinating aspect of their findings was about how listening habits changed during the pandemic. “We did a comparison between South Africa and Australia,” says Askin. “In South Africa, which had much stricter lockdowns than Australia, people were more likely to explore new musical styles. Both had some restrictions, but the countries with stricter lockdowns were much more likely to be more exploratory.” The researchers attribute this to greater disruptions in people’s everyday existence.
Proactive Versus Reactive Consumption Habits
Askin and his coauthors also realized people listen to music and explore cultures more generally to provide a soundtrack to their lives rather than as a passive act. “Some people have a folk theory that we let the world around us dictate what we consume, as opposed to being more proactive and exploratory, but our research flips that idea on its head and moves it in the other direction,” he says. “People are much more proactive in terms of their cultural consumption.” We seek more interesting, novel, and exploratory forms of art as we do more novel and exploratory activities.
The Importance of Retooling Algorithms for Disruption Patterns
Askin is excited about the implications of their research for the future. “Obviously Spotify, Deezer, Apple Music, Tidal and all of the other streaming services already know where you’re listening and what devices you’re listening on,” he says. “But understanding how to tweak recommendation algorithms to acknowledge the differences people experience—then match accordingly—is critical. It’s not just about how to get people to listen to new songs we’re sure they’ll like. There’s an opportunity as to when to introduce the new music, not just about what songs to introduce. Ultimately, it’s about a deeper and better understanding of taste development and taste evolution over time.”
Read the research article here: https://www.pnas.org/doi/10.1073/pnas.2306549121
Associate Director of Communications
jrotheku@uci.edu