You can find this website, which I’m currently updating, at brycedemos.com.
In early 2020, I was working on a small Python experiment trying out Spotify’s free API. I learned that Spotify uses AI to measure abstract attributes of a song such as acousticness, mood, and energy in order to recommend songs to users. Immediately, I wanted to visualize some selected attributes in three dimensions instead of a list of numbers. As I improved the visualization and moved it to a web experience using Django, I created Musicscape with the intention for people to visualize their own music tastes in a spatial environment.
A need for algorithm transparency
I put the project down, and a few months later I was surprised to find there were way more users than I expected from Eastern Europe as well as some blogs that wrote about Musicscape. It seems like there is a common thread among online discussions: Musicscape helps people understand otherwise hidden aspects of Spotify’s recommendation algorithms. For example, Gargantua wrote:
“Якщо вам завжди було цікаво, як Spotify створює плейлисти, то цей ресурс — для вас. Сайт brycedemos.com візуалізує алгоритми сервісу та пояснює користувачам, як стрімінґ-платформа формує особисті бібліотеки.”
(English translation)
“If you’ve always wondered how Spotify creates playlists, this resource is for you. The site brycedemos.com visualizes the algorithms of the service and explains to users how the streaming platform forms personal libraries.”
A music blog, Slukh, wrote:
Усі гадали, як Spotify створює такі класні плейлісти. Відповідь дав сайт brycedemos.com: розробники взяли відкритий API та використали параметри, які стрімінг-сервіс застосовує при створенні рекомендацій.
(English translation)
Everyone wonders how Spotify creates such cool playlists. The answer was given by the site brycedemos.com: the developers took an open API and used the parameters that the streaming service uses when creating recommendations.
Thanks to these blogs, they helped me understand that most of my users are curious about using Musicscape to understand Spotify’s recommendation algorithm more than understanding their own music tastes.
Some takeaways:
I didn’t do any promotion of my website, and I learned that Musicscape spread online with the word of mouth from “advocates”. Looking at the statistics of use, there were commonly users that brought their friends to both share their music tastes and understand Spotify’s recommendation algorithm. This is similar to how Spotify Wrapped is shared among circles of friends every year.
More importantly, In the wake of Cambridge Analytica and similar events concerning data, many people are curious about what data is being collected and how it’s being processed by companies like Spotify. Here are some popular memes poking fun at Spotify’s algorithm. They’re jokes, of course, but they reflect curiosities and questions people have.
