Even off the top of the charts, this seems to happen all the time on Spotify on their automatically-generated playlists, especially the "song radio" lists. It plays Lucinda Williams's "Fruits of My Labor" for me all the time, which is a good song – I'm not gonna skip it but, you know, I like a lot of other songs by her, too. That song has nearly twice as many streams as her next most popular one, which just doesn't track for me. It seems like maybe their algorithm becomes self-fulfilling.
And then you get stuck with something you hate. For a while it kept feeding me Wayne Shorter's "Porta de Areia." I love Shorter, but it's not typical for him and it's definitely the kind of song where if you don't like it, you really really hate it. It's taken me a long time to convince Spotify not to play it for me.
This is such a thoughtful analysiss of algorithmic bias. The comparison between Spotify and Apple Music's top songs reveals something fascinating - 'Sweater Weather' at #10 on Spotify vs #154 on Apple is a massive disparity for what should be a universally consumed platform hit. Your point about 'million small reasons' rather than deliberate manipulation feels right - these systems compound tiny preferences over billions of interactions. What really struck me was the Arctic Monkeys example: 2.4 billion streams on Spotify but barely 100 million views on YouTube. That's a 24x difference for the same song, which suggests the platform isn't just reflecting user preferences but actively shaping them. The observation about Spotify favoring 'laidback indie or alternative rock that won't disturb you in the background' makes me wonder if this is algorithmic convergence - the system optimizing for 'safe' recommendations that minimize skip rates rather than maximizing engagement. Do you think we'll see these biases become more transparent as platforms mature, or will they remain effectively invisible to most listeners?
I just played Sweater Weather on YouTube. I don't think I've heard that song before. It's so weird when a song can be so popular, yet I miss it entirely.
Even off the top of the charts, this seems to happen all the time on Spotify on their automatically-generated playlists, especially the "song radio" lists. It plays Lucinda Williams's "Fruits of My Labor" for me all the time, which is a good song – I'm not gonna skip it but, you know, I like a lot of other songs by her, too. That song has nearly twice as many streams as her next most popular one, which just doesn't track for me. It seems like maybe their algorithm becomes self-fulfilling.
And then you get stuck with something you hate. For a while it kept feeding me Wayne Shorter's "Porta de Areia." I love Shorter, but it's not typical for him and it's definitely the kind of song where if you don't like it, you really really hate it. It's taken me a long time to convince Spotify not to play it for me.
Fascinating read. Just ordered the book. Great work, thank you @chris
Thank you!
This is such a thoughtful analysiss of algorithmic bias. The comparison between Spotify and Apple Music's top songs reveals something fascinating - 'Sweater Weather' at #10 on Spotify vs #154 on Apple is a massive disparity for what should be a universally consumed platform hit. Your point about 'million small reasons' rather than deliberate manipulation feels right - these systems compound tiny preferences over billions of interactions. What really struck me was the Arctic Monkeys example: 2.4 billion streams on Spotify but barely 100 million views on YouTube. That's a 24x difference for the same song, which suggests the platform isn't just reflecting user preferences but actively shaping them. The observation about Spotify favoring 'laidback indie or alternative rock that won't disturb you in the background' makes me wonder if this is algorithmic convergence - the system optimizing for 'safe' recommendations that minimize skip rates rather than maximizing engagement. Do you think we'll see these biases become more transparent as platforms mature, or will they remain effectively invisible to most listeners?
I feel like most people listen to music passively, so won’t notice unless recommendations become tremendously bad
“Bless the Telephone” 😭
I just played Sweater Weather on YouTube. I don't think I've heard that song before. It's so weird when a song can be so popular, yet I miss it entirely.
I just searched "505" on YouTube and it has 695 mln views.
That’s odd. Link me. Still significantly fewer streams than Spotify.
https://www.youtube.com/watch?v=qU9mHegkTc4
Thanks. It’s says it’s unavailable for me. Guess that’s why it didn’t appeal. I’ll try to update with another example.