More like an alNOrithm

At some level, most of us interact with discovery algorithms on a daily basis. They take the form of “new music for you” playlists, “Picks for you” movies, or “Recommended” books. These tastemaker algorithms are supposed to look at what you enjoy, then recommend new and exciting things to you. But that’s not how taste works, and this loses the magical sense of discovery when you find your new favorite thing.

I tried tackling this problem a couple years ago at The New York Times, but while I was trying extremely hard not to come off as a complete idiot in the op-ed section and adding lots of links and further reading, I missed the point: finding new media that you love, on your own, is a joyful experience. It’s fun. Who the fuck cares if it’s anything more than that? Isn’t the pursuit of fun things that feel nice enough?

Finding new and interesting stuff isn’t easy. It was never easy. It perhaps never will be easy, because when companies try to make it easy, it stinks. A company has profit models in mind. It wants engagement and retention. We just want something cool.

We are also complicated. Most people do not like just one genre of music, movies, or books. We do not all want the feedback loop of reading every book by the same author, or listening to every album—and at this point, every deluxe album, reissue, special edition, uncut track, and whatever else—or watch movies in a single genre. That’s just not how most people work, despite that being the only thing that algorithms can help with.

Let’s just take music as an example, since Spotify has one of the most ouroboros  recommendation algorithms of any service I’ve ever used. Let’s say you’re a fan of Joyce Manor and you listen to a lot of it. Spotify recommends more Joyce Manor, and more emo, usually of the same era. It’s as simple as that. You can accomplish the same thing by spending five seconds searching for “bands like Joyce Manor.” And besides, most of the time I don’t want to just get a recommendation for a band that sounds like the band I just listened to. I already have that band, I don’t need another.

This is the same with any sort of taste curation. It just feeds back the same things, over and over. This can be useful, in a way—say you like Hot Snakes and Spotify recommends Drive Like Jehu and Rocket From the Crypt, three bands that share members—but that was rarely the case in my experience.

But again, I’m over here trying to justify that the technology sucks by using examples. Which isn’t really the point. When my laptop says, “Dude, if you like Hot Snakes, you might like Rocket from the Crypt,” it means nothing to me in the moment. I’ll forget that moment in an instant. In reverse, I distinctly remember the day I walked into high school with a copy of Rocket from the Crypt’s Scream Dracula Scream CD and my cool, older, punk rock mentor told me, “Dude, get ready to have your mind blown by Drive Like Jehu.” I used to find new music through a combination of talking to friends, reading zines, and just walking aimlessly through record stores.

I remember the day I was walking through Tattered Cover bookstore in Denver, Colorado, trying to find something new to read while avoiding a gnarly incoming snow storm at some point in 2004 or 2005. I walked up and down the aisles, picking up books and staring at the back covers, until I landed on David Mitchell’s Cloud Atlas. It had a sticker on it that said “Recent finalist for a Man Booker prize,” and was on a table with other recent finalists. Nothing else on the table really spoke to me, and frankly, I had no idea what the Man Booker prize even was, but Cloud Atlas sounded interesting enough, so I purchased it. It became one of my favorite books of all time.

Can you recreate this sense of discovery digitally? Of course. Can you feed enough information into a store to get back something useful sometimes? Sure. But if the things we end up loving the most in the longterm are the things that surprise us, algorithmic discovery will never get us there. I know I sound like a cranky old dude, but whatever, I’m a cranky old dude.

So, okay—maybe by some off chance you’re still with me. Your eyes didn’t roll all the way back in your head and you’re asking, “well what the fuck am I supposed to do then, Thorin?”

Here’s what I’ve done, ever since writing that article in the Times.

  • Switched from Spotify to Apple Music. I’m not sure if it’s intentional or not, but Apple Music’s algorithmic discovery is terrible. So I never use it. As far as I can tell, it doesn’t create weekly or daily playlists like Spotify does either, but that could also just be the wonky design of Apple Music. In any case, whatever it does do, isn’t shoved down your throat like Spotify.
  • I unfollowed nearly every political journalist on Twitter (sorry) and replaced them with music writers across genres and publications. I’ve signed up for newsletters from a couple of them—I like NPR reporter Lars Gotrich’s “Vikings Playlist”—and listen to  the occasional music podcast, like Indiecast. I peek in on sites like Stereogum, Brooklyn Vegan, Pitchfork, and whatever else I’m not thinking of right now.
  • I signed up for Letterboxd, then followed a handful of cool sounding people/critics, so I’m drip fed a series of new and old movies that sound interesting. This has been considerably better than any other movie discovery tool I’ve used, since it’s just real ass people with normal opinions that you get to know over time. Did I want to sign up for yet another social network? Absolutely not. But I’m glad I did.
  • I started keeping a spreadsheet of books to read, but that quickly grew a bit unwieldy over time and I wasn’t doing a very good job of maintaining it. I then tried a to-do list, which lacked the context that I needed (why did I add this book to my list????), but eventually came across an app, called Sofa, that’s built just for tracking these types of things (it can also track movies, video games, and more—I just use it for books and TV shows). It’s great. I ended up paying for the yearly plan, even though I’m not totally sure I need most of the features. It’s also nice because you can export to a CSV file, so if I decide to stop using it or whatever, I can return to my spreadsheet approach.
  • As for finding books to read, I have this penchant to search for extremely specific terms to find books that I’m in the mood for, where I’ll go hunting for like, “epistemological experimental science fiction novel,” or whatever. So I tend to build up lists that way. I also keep an eye on new releases, read sites like Electric Literature and the NYT Review of Books (which has great little mini lists pretty often, even if they tend to focus longer article on books I’m less likely to be interested in). I’m also a massive fan of NPR’s Book Concierge, which they release every year around the beginning of December. The Book Concierge tasks its numerous critics to pick books in a wide variety of genres from a variety of voices. I love it, and look forward to it every year. I always end up with at least a dozen new books to read.

I think, that’s mostly it? I guess I’m just naturally curious, so if I see a concert poster (or when I used to, anyway, pre-COVID) that looked cool, or a random book ad or whatever, I’d look into it. I still do that, though I feel like the various states of lockdown over the last two years have hampered that accidental discovery a bit. I’ve been trying to be more active with engaging with friends, even just over text messaging, about new bands or movies I like, just like we’d be doing if we were in the same room. It works out, occasionally, and I’ve found some great stuff.