The surprising thing I learned from quitting Spotify

Last summer, I quit Spotify, and wrote about it with the rather unsubtle headline “Why I quit Spotify.” My reasons remain sound: The software had become clunky, the ads relentless, and the Sabrina Carpenter songs too inescapable. I wanted to find a better music streaming service. It gives me no pleasure to report that a few weeks ago, I rejoined.

The algorithm got me. I don’t just mean that it got me, the way the TikTok algorithm glues you to the screen. Spotify’s algorithm got me the way an old friend gets me and my weird affection for yacht rock or ongoing obsession with French touch music from the mid-Aughts. It took a few months of digging through the proverbial crates of Apple Music for me to realize that Spotify has something other streaming services could never get: 15 years of my music listening habits and artificially intelligent software to reinforce those habits.

This is why algorithms tend to be viewed as villains these days. They’re the technology behind TikTok’s For You page, which keeps feeding you weird videos you can’t stop watching, and Amazon recommendations that appear to know what prescription you’re taking. Facebook’s algorithms, meanwhile, have been radicalizing Americans for at least a decade, and Instagram’s algorithmic feed is wrecking the mental health of an entire generation. The implications of Spotify’s algorithms, you could argue, are quaint by comparison.

Quitting and unquitting Spotify made me realize something, though. As central as algorithmic feeds are to how you consume information, you have more control over how those algorithms shape your tastes and behavior than you might think.

If an algorithm works for you — as Spotify’s does for me — don’t feel bad about submitting to its effortless and convenient offerings.

Music has always been important to me, and over the years, it started to feel like I had to gamify Spotify to find songs that I truly loved. When Spotify launched in 2011, it was basically a massive library of all the music, but over the years, it introduced more and more algorithmic recommendations and playlists that promised to match my taste. It still took work to find the good stuff.

This work is what has now made Spotify’s algorithms irreplaceable to me. It has a decade-and-a-half of my listening history, and over the years, I’ve learned its quirks and tinkered with it to meet my needs. I spent months trying to replicate this experience on Apple Music, but its algorithms struggled to surprise me.

All music streaming algorithms operate on two basic principles: content-based filtering and collaborative filtering. The content-based filtering tries to identify specific aspects of a song itself, including the artist, genre, mood, and so forth, to queue up the next song. Collaborative filtering refers to recommendations made based on other people who listen to a certain song and what else they listen to. If two people listen to the same five songs, there’s a good chance they’ll both like this sixth song. It’s all math, and sometimes there are anomalies that will delight you.

“Some of the serendipity that you get is sort of error turned into virtue,” Glenn McDonald, a former data alchemist at Spotify and creator of Every Noise at Once, told me. “So you’re surprised, and sometimes those surprises are pleasant.”

It’s not just that Spotify’s recommendations tend to be pleasant because it has a lot of data about me. It’s that Spotify has the listening history of 675 million people, whose interests may overlap with mine in countless different ways. Over the years, I’ve developed a set of habits that help me hone those recommendations — things like making playlists, rejecting recommendations I don’t like, exploring artists’ catalogs, and maybe most importantly, digging through other people’s playlists.

This is what I call lean-forward listening. While it’s easy enough to click on Discover Weekly every Monday, lean back and listen to the whole thing like a radio show, and then move on to the next playlist, the more effort you put into curating your experience, the better the algorithms will work next time. At the very least, you’ll find your way onto a playlist that algorithms didn’t create.

Like them or not, algorithmic recommendations aren’t going anywhere. Companies like Spotify like them because — when they work — algorithms keep people hooked on their products. Companies like Amazon like them because algorithmic recommendations enable them to steer people’s behavior. The right product recommendation could lead someone to buy something they didn’t otherwise plan on buying. (We’ve all done it.)

This status quo seems dystopian in a lot of ways. Algorithmic recommendations were all the rage a couple of decades ago, when personalization felt convenient rather than creepy. Netflix deserves a lot of credit for this, since it pioneered the concept of giving you customized movie recommendations in the late 1990s. But by the early 2010s, it was getting hard to tell the difference between personalized recommendations and targeted ads. Now, practically everything you see online is personalized to a degree, from the front page of the New York Times to the list of restaurants in your favorite food delivery app.

You can probably learn to live with it when you’re talking about music on Spotify or burrito restaurants on DoorDash. “The stakes are a little bit higher when it comes to recommending things like products on Amazon, and even higher when it comes to recommending things like content on Facebook,” said Meredith Broussard, a data journalism professor at New York University. “Because, as we all know, disinformation and misinformation are very, very popular, but not good.”

The role algorithms, which are designed to boost engagement, play in spreading misinformation is a book-length topic. For now, I’ll just reiterate that you don’t have to lean back and let Facebook, Google, or X flood you with algorithmically generated information. You can learn more about how these platforms use algorithms and steer them to your advantage.

If you’re sick of the algorithm on X feeding you right-wing propaganda, try Bluesky, which lets you pick different algorithms for your feed. And if Netflix or any other streaming service has gotten stale, try nuking your view history and starting over. Spotify offers a list of details about how it recommends content and how you can make tweaks. And Amazon has a tool that’s designed to improve your recommendations. (I have tried all of these things, including the Amazon tool, which is very tedious but still possibly helpful.)

Things get a little tougher on big platforms like Google, Facebook, Instagram, and TikTok, whose algorithms tend toward the black box end of the spectrum. Still, knowing how algorithms work and playing an active role in making them work better for you can improve your experience on almost any platform. Algorithms are only in charge if you let them be.

In some cases, you might like it when the algorithm’s in charge. This is how I generally feel on Spotify, although I’m constantly correcting it and guiding it. This is also how I generally feel on Amazon, where I try to buy only the basics. I quit Instagram a while ago when I decided the algorithm was in charge a little too much. If I get bored one day, I might try it again.

A version of this story was also published in the User Friendly newsletter. Sign up here so you don’t miss the next one!