Undergrad's Research Informed by Your Mood and Playlist

Spotifeel is a new computer program to help scientists evaluate the relationship between your mood and your playlist, developed by NJIT rising senior Rain Nutt for her undergraduate summer research.

Existing efforts tend to evaluate all listeners by the same set of music, whether they like the songs or not, so Nutt decided to gauge them on their own choices. Any loss of objectivity in a study would be countered by an increase in the test subject’s comfort level, she asserts. “You can work from where they are, instead of them coming to you,” she stated.

The method gauges a Spotify user’s 25 top monthly songs based on categories that the music service calls acousticness, danceability, energy, liveness, speechiness and valence (positivity). Nutt wrote her code in Python and was advised by David Rothenberg, distinguished professor of humanities and social sciences, after taking his course in electronic music.

"Rain has proposed a serious way to personalize Spotify's algorithms so that you can ask your own questions about the music it picks for you, not just the questions the company wants to ask. It's a step in the right direction for music recommendation engines," Rothenberg noted.

“Spotify analyzes every song in its database algorithmically. It does a lot of different predictions, but it's essentially algorithmic predictions of various qualities of a song. It's one of the things that they use to construct recommendations and identify music that's similar to each other. Because I knew that they analyzed every song that way, it struck me as a way to analyze music, the relationship between you and the music you listen to, in a way that was a lot more organic,” Nutt explained.

A caveat of Spotifeel is that it doesn’t analyze lyrics, so it would not know the difference between similar-sounding songs that have uplifting versus dark words, Nutt acknowledged. Bruce Springsteen’s “Adam Raised a Cain” is musically similar to “Prove it All Night” from the same album, although the former laments a troubled parent-son relationship and the latter celebrates youthful determination.

Nutt was unable to actually perform a survey due to time constraints -- her assignment was to develop the method -- but she hopes other researchers will integrate her open-source code into their work. Perhaps most importantly, she learned a lot while designing Spotifeel.

“It was helpful in learning a lot about how to interface with existing APIs, how to deal with API keys and do authentication and that kind of thing, which was mostly new to me. But it is knowledge that can be pretty well generalized to other websites and other APIs,” Nutt explained. “I also feel like I have a deeper understanding of what Spotify looks like in a more honest sense, as opposed to the things that you see, the recommendations that you see. It obviously doesn't say you're recommending this song because it has these similar scores. It just kind of shows it to you. And so working with these numbers more directly, being the one who pulls from Spotify, sends them songs and gets these results back, it makes me a lot more familiar and starts illustrating themes in how both Spotify works and how things that use Spotify work. It was having that hands-on experience that allows you to be a lot more perceptive about what Spotify actually means and what it does, instead of just being the machine that dispenses music.”

Looking forward, Nutt said she’d like to make the program more modular, easier to use and have it produce data accessible to the public. She is also thinking about what to do after graduation, which might include a career in user interface design or perhaps graduate school. Working for Spotify itself would be a dream job, she said.

“I appreciated having the opportunity for the summer research. It’s a way to demo research and feel out what it's like and how things might snag, or how to work with other people, how to present things, and it feels like something that I do really see a future in,” she concluded.