A particle physicist. A social psychologist. A musician-turned-computational neuroscientist. What are these wide-ranging backgrounds doing together at Spotify? They’re dedicated to understanding people through music, a research initiative that has become a key part of Spotify’s data mission. The theory behind their work: because music listening is so uniquely emotional, universal and, now, addressable thanks to streaming, it can uncover deeper insights than consumption of other kinds of content like movies and TV.
“There are very few other forms of content that are small, digestible, that people repeat a lot,” says Clay Gibson, product owner. “And because everybody has a different music taste, what you listen to is tied to who you are as a person, how you feel, where you’ve been, and what you’ve experienced. Given that people are engaged with Spotify throughout the day, we have a wide window of observation to start to understand these things.”
The initiative kicked off in January of 2016, when Spotify’s VP of Data, Adam Bly, determined that a new approach to studying Spotify’s data could yield a more holistic view of who people are. From there, a team of researchers took on the analysis, which started out by identifying attributes unique to music consumption through streaming that they could use to observe data patterns over time. They’ve focused on three dimensions so far: Discovery, Diversity and Tilt, which measure how much people seek out unfamiliar music, explore a wide range of sounds, and actively curate their listening, respectively. Several more are currently in exploration, and the team is readying a research paper to share their earliest findings. In June, Bly will also share some of these findings onstage at Cannes Lions.
“We’re creating these metrics that describe users and can use those to try to predict these stable, off-app, external traits,” says Gibson. “If we know about someone’s Tilt, Diversity and Discovery, do we know if they are introverted or extroverted? Do we know if they like cats? Or horror movies? Or video games?”
Even with their deeply scientific backgrounds, the researchers say Spotify’s dataset is extraordinarily rich. “With traditional psychological research, you’re asking people questions and they’re answering on a scale from one to five, and those answers are used to get some sort of universal truth about them,” says research scientist Scott Wolf, who studied personality prediction before tackling this project. “With Spotify’s data, it’s people acting like they would normally, day in and day out. They’re not trying to present themselves as something they’re not. So that allows us to get a more true picture of a person, rather than what somebody wants to present themselves as.”
For Alice Wang, who joined the team in May, using science to predict behaviors is nothing new. Wang spent years studying the same exact traits in the brains of mice navigating a VR maze, and before that she was a violinist for the art-rock band Ava Luna. It’s a varied background, and one that’s perfectly suited for her new gig.
“Music is one of the few activities we do in life that engages all parts of the brain—emotional, physical and cognitive—all at once,” Wang says. “I fundamentally believe that people’s musical taste provides a window into understanding their soul.”