In this episode of the Arts Management and Technology Lab, Samantha Childers explores why so many songs recommended by streaming platforms like Spotify, Apple Music, Amazon Music, and YouTube Music sound strikingly similar. She breaks down the inner workings of music streaming algorithms—how metadata, collaborative filtering, content-based filtering, and continuous feedback loops collectively shape the “perfect playlist.” Samantha also examines Spotify features such as Discover Weekly and Discovery Mode, uncovering how these tools influence music discovery, artist visibility, and even compositional trends as musicians increasingly write with algorithms in mind. Alongside these technical insights, she raises critical ethical questions about data privacy, algorithmic bias, pay-to-play practices, and the shrinking role of human curation. Ultimately, the episode invites listeners to rethink how algorithm-driven platforms shape not only what we hear but also how we discover, value, and experience music.
Show Notes
A Critical Research of Spotify’s Business Model—The Case of Discover Weekly
We Tested Popular Music Streaming Services — These Are the Top 6 You Should Check Out In 2025
The Role Of Music Streaming Algorithms In The Industry
Algorithmic Symphonies: How Spotify Strikes the Right Chord
Algorithms in Music: Blessing or Curse?
Understanding recommendations on Spotify
How to Break Free of Spotify’s Algorithm
Pay-to-Playlist: The Commerce of Music Streaming
Impacts of AI on Music Consumption and Fairness
Transcript
Samantha Childers
Everything is moving more towards convenience, and this has really affected how musicians are composing as well. The lengths of their songs, the catchiness, they're essentially changing the way they write to write for the algorithm.
Dr. Brett Crawford
Welcome to another episode of Tech and the Arts. The podcast series of the Arts Management and Technology Laboratory, also known as AMT Lab.
The goal of our podcast series is to exchange ideas, discuss emerging technologies, and uncover ideas that arts managers and arts geeks like us need to know. My name is Brett Ashley Crawford. I'm the executive director and publisher for AMT Lab. In this episode, Samantha Childers talks about her deep dive into the world of music algorithms.
Show notes, related links and a full transcript are available on our website, amtlab.org/podcasts-interviews. With that, here's Samantha.
Samantha Childers
As a musician, I have spent a lot of time listening to music and exploring various music streaming platforms. Mostly the free versions because you know, I'm a musician educator and we all tend to make a lot of money.
I've tried most of the music streaming services, specifically remembering Pandora, sticking out to me when I was younger, but feeling like I heard the exact same songs over and over again. Expanding to Spotify's free version during my undergraduate and graduate degrees, feeling a little bit better that I was hearing at least different songs, but they all still sounded the same.
Then getting access to a free three month trial of Apple Music, being really impressed by the different playlists they had. I was hearing new songs and new musicians, but again, the sounds were still the same, similar vibe, similar sound. I've always wondered why all this music I was listening to sounded the same. Was it the technology of Pandora just not being up to date yet, back in the two thousands, 2000 tens, was it because of paywall and I wasn't paying for these services? Was it the way that I was searching for playlists by searching just for an artist, or was there another factor at play? And now that I've gone down the rabbit hole, I've confirmed this.
This is because of the way that music streaming services recommend songs and playlists via algorithms. And I've confirmed my suspicions, they are pigeonholing our taste in music. So starting with the very basics of music streaming services, according to a 2023 research article by Wang, the four most prevalent music streaming services are Spotify, Apple Music, Amazon Prime Music, and YouTube Music.
Each of these have different levels and features, including prices and types of sound resolution and their libraries. An article that I found from the Rolling Stones confirmed that typically Spotify comes out on top. On almost every rating, which is curious because it's actually the most expensive at $11.99 for a regular single user account.
It does not have the most advanced audio quality Dolby Atmos sound, lossless audio. Whereas Apple, YouTube, Amazon, they do, and the Spotify music library only has 50 million songs, whereas Apple Music has 100 million. So it's interesting to me that Spotify is the most popular, but what Spotify does have is a free tier and an advanced algorithm recommendation that gets better as you use it and as you sign up for a premium subscription. Spotify reports in a newsroom press release and that they have over 696 million users worldwide. And only a third of those users are subscribers. So the appeal of this free tier is that you get access to this 50 million song library, but you're forced to sit through ads not having any skips and less personalized playlists, which Wang describes and explores as a freemium business model.
But Spotify's big value proposition is its algorithmic recommendations that help create these customized and personalized playlists ranging from a specific artist all the way to a feeling or a vibe. This is really effective in converting users on that free tier to a subscription. So how do these algorithms work?
I consulted three different sources to get a full understanding of this. A Booth collective blog, a USC journal and a 2018 thesis from SMU as well as Spotify's method that they describe on their website. So essentially it all starts with data collection on songs and users. Metadata is collected on the songs, which is based on audio data like tempo, key beats per minute, artist data genre and natural language processing technology can be used to scrape the web for that kind of stuff too.
We then get data collected on the user and their habits, including demographic information like age or even location, and how users are interacting with the platform, likes, skips, searches time, listening to a specific song. Once this data is collected, we move to a filtering model, mainly through two different methods, collaborative filtering and content filtering.
As described in Andrew Penney's, SMU Thesis, collaborative filtering, a quote here, “looks for users with similar preferences and recommends songs or artists that those similar users have enjoyed, whereas content filtering analyzes the audio features of the song such as tempo, key instrumentation and style to identify similarities and recommend songs that have similar attributes to those a user has enjoyed in the past”.
So collaborative, looking at users and content, looking at the song and comparing. Finding those similarities, ultimately recommending songs that you're going to like based on similarities of the songs you liked. A, so try listening to B, or similarities between different users. This person listened to the same song as you, and they really liked this.
Why don't you give this a try? You're likely going to like it. All of these tasks can be supported through machine learning, which is looking for patterns as well as feedback loops that are tracking how users are interacting with their software. And the algorithm will then adjust, and this is specifically from Spotify's website.
They say your recommendations are constantly influenced by your engagement with content on Spotify. The more you listen to content you like and the more you interact with the app, the more we think you'll enjoy your recommendations. Spotify does this super, super well through personalized playlists, which again can be based on an artist, a mood, a time of day, a general vibe.
They have this market down and other music streaming softwares have followed Spotify in this playlist descriptive vibe. I had a friend that got a Spotify playlist come up for her called Sad Monday Cry Sesh or something along those lines, like it's pretty effective. Glen McDonald, who is a former engineer at Spotify, explains that these playlists essentially fall into three categories.
We've got the stuff we like, the stuff like we like, and then things that are completely from left field. But usually we're gonna stay in those first two camps. And this ultimately results in homogenous sounding playlists. So confirming these experiences I had from the very beginning. For example, I know that when I put on a Lizzie McAlpine playlist, I'm probably gonna hear Gracie Abrams and Noah Kahn and Hozier or songs or musicians like that.
And this is what can be called filter bubbles confirmed by our USC journal article where they are quoting that users are only recommended music that aligns with their existing preferences, limiting exposure to new genres. So that's our filter bubble, which I think is a pretty good definition for what we are experiencing.
Then we have the efforts from music streaming services that help users discover new music. So again, looking at Spotify, their main two are Discover Weekly playlists and discovery mode. And these can help increase the variety of music and artists that users are exposed to. But keeping in mind that these music streaming services are businesses and are focused on revenue, so they wanna keep users happy and not feed them too many things out of left field.
So Discover Weekly was launched in 2015 and essentially provides users a new playlist every week on Monday of 30 tracks that are aimed at helping listeners discover new and smaller and niche artists. Now in order for those smaller, newer niche artists to get onto those playlists, they often accept a lower royalty or a lower than market rate in order to get themselves there.
This brings up a pay-to-play concept, which is frequently referred to as a Payola. I think I'm pronouncing that right, but in this situation, it's a reverse payola. They're not paying to be played. They're accepting lower compensation just to get placed onto this. List this playlist, and this is explained further in a 22 journal article by Christopher Jay Buca, Fusco and Christ Garcia.
Looking now at discovery mode. This is a tool both for listeners as well as artists. So for listeners, you can turn on discovery mode in Spotify again, which changes your algorithm to suggest songs that you haven't heard before. More diverse music. It's essentially saying like opening a door and saying, hey, I'm opening stuff to stuff that doesn't sound like my stuff. Now looking at it from the artist's perspective, and this is a quote from Spotify's website, “Discovery Mode gives artists and labels the opportunity to identify songs that are a priority for them, and our system will add that signal to the algorithms that determine the content personalized listening sessions. When an artist or label turns on discovery mode for a song, Spotify charges a commission on streams of that song in areas of the platform where discovery mode is active”. Even still, artists have to meet a certain criteria to be eligible to even use discovery mode. So thinking about the artists who don't meet that criteria, how do they even get streams?
And then there's that ever prominent ever present fact in the back of our brains are like, wait, what about my data? We forgot about that. How is Spotify using it? All of these are really big topics of debate right now of just the ethics behind music, streaming services, algorithms, and generally just the biases that are inherently programmed into these tools and how our own biases are affecting the music that we want to listen to.
And keeping that in mind that music streaming services are tending to promote music musicians and artists from larger labels like Sony and Universal. So how fair really is that? There's an interesting solution that I read about that concerns like how playlists are created, and it's called the Fair Muse Project.
And this is specifically a project that's happening in Europe, but it boiled down to its simplest description. This project essentially places a score on a playlist on how fairly it was made. So the solution isn't trying to change the algorithm or change the playlist, but instead educate listeners on how fairly the music that they're listening to, how fairly that playlist was created.
I thought that was a really interesting solution or an interesting take at least. So ultimately the way that we explore and listen to music has completely changed. Thinking about gatekeeping, which was discussed by Tiziano Bonini and Alessandro Gandini in their article. So we're moving away from human curators and editors and much more towards this algorithmic recommendation.
We've we're moving away from that social interaction that music listening and music exploration often had, which is reinforced by Tiffany's MIT article. Just thinking back to when people made mix tapes or burned CDs of their favorite music and shared it with their friends, that's almost gone unless it is December and Spotify Wrap has come out and everyone's gonna post that on their Instagram story.
Everything is moving more towards convenience. This has really affected how musicians are composing as well. The lengths of their songs, the catchiness, they're essentially changing the way they write to write for the algorithm. So I think it's important for us to think about the role that music plays in your life and how you are wanting to experience it.
Do you want to have a comforting and familiar music on in the background? Then an algorithmic recommendation playlist might be right for you. That's totally fine. These music streaming services are gonna be your friend. They're happy to do that for you and happy to take your money. But if you want to explore and experience new and diverse music, the top four might not be for you.
And different methods of listening to music might be better, like curating your own playlists or exploring other music streaming softwares out there. Thank you so much for listening. I had a really interesting time exploring and confirming and validating myself that music streaming softwares definitely pigeonholes you into listening to certain types of music. They wanna keep you happy so that they'll keep you engaged.
Dr. Brett Crawford
Thank you for listening to this episode of Tech in the Arts. If you found this episode to be informative, educational, or inspirational, be sure to check out our other episodes and send this to another arts or technology aficionado in your life. If you want to know more about arts management and technology, check out our website at amt-lab.org, or you can email us at info@amt-lab.org. You can follow us on Instagram at techinthearts, or on Facebook and LinkedIn at our full name Arts Management and Technology Lab.
Our season runs from September through May, so feel free to follow us on your preferred podcast app, be it iTunes, Spotify, or YouTube.