Against Discovery
breaking up with the perpetual recommendation machine
Hey! It’s been really gorgeous getting to connect with you all over the winding course of this newsletter. So many people who work as managers, as artists, in tech, for Spotify (omg drama), have read and engaged with my work. I’m so thankful. If something I’ve written has sparked something in you, and if you would like to chat about how the themes I write about pertain to your work, drop me a line. I’d love to know, firsthand, how you (the manager, the artist, A&R, the Spotify employee) are dealing with the algorithmic mind melt.
You also can just say hi :)
My email is tobiehess@gmail.com
Please though subscribe to this newsletter if you have not yet!
Also, I do this for work free. Right now, it’s a pure passion project. If my writing has resonated with you before, please consider leaving a tip to support this work. Doing so helps me dedicate more time to this labor of love. Who knows, maybe soon it can be my full time gig? But for now, a tip jar is what I have. Thanks!
I’ve seen the greatest minds of my generation devote their waking hours to sorting through the muck and shit.
Past generations tackled such gargantuan challenges as perpetual motion and food production, but now, if you’re bushy-tailed and blessed with a top-40 college degree, there’s a good chance you have landed yourself a cushy gig working some sort of sorting project. What do we do with all of this content? What do we recommend to the individuated masses? It’s a problem as urgent as nuclear fusion.
Discovery, like “disruption,” is one of those tech-culture concepts so constantly invoked that it seems to turn to mist as soon as it’s spoken. Aw yes, discovery . . . like the channel, no! . . . like Columbus — GOD NO . . . okay, like stumbling into a whole-in-the-wall taco joint??? I mean sort of. In tech-land, “discovery” seems to invoke a low-key unearthing surprise. For example, Discover Weekly, the algorithmically-curated playlist given to Spotify users weekly, is the platform’s way of gifting each of us our very own box of chocolates. Based on an analysis of our tastes and habits, we receive these 30 gems. The limited-scope of that offering positions discovery as a discrete, describable event. We come to the playlist to discover.
But the platform more broadly displays a different exploratory ethos.
I have written ad nauseam about Spotify’s unstable, constantly-shifting listening experience, how their prioritization of background music and their constant push to “recommend” music makes it hard for any of us to develop a stable musical taste and identity. Despite such discovery-events as Discover Weekly, the notion that discovery is perpetual is baked into the Spotify experience. For years, you would open the app to a few albums you’ve “been listening to,” but also, a playlist titled “fresh new indie,” and under it: a banner titled “vibes,” under of which is a vaguely determined selection of brand new picks. The whole home page is a complex curatorial mood board.
When recommendations are basically all you see, discovery becomes less a glorious, one-off encounter and more akin to cutting through the untamed bramble. Cut. Cut. Cut.
And now, this discovery process has gotten a face lift.
A few weeks ago, Spotify launched their new TikTok-style vertical feed where users can sift through recommendations in one clean, Gen Z swipe. This will be an integration of musical recommendations, yes, but also audiobooks, podcasts, and even video.
On the top of my #NewFeed is a mosaic of 8 possible vibes-based mixes. Some are themed by era, others by genre, and some by artist. Then underneath this wall of wonders is Spotify’s new AI DJ . . . which I can’t even begin to talk about right now. Like. Watch the video and idk, stay tuned for my upcoming thoughts. And then, well, we see Tha Feed! Different from the complex assemblage that was their home screen before, this new feed is infinite, streamlined, responsive to our changing tastes and habits. And now, rather than going through the labor of sifting through a cluttered homepage, you can simply glide along their track.
On the feed, each swipe-down brings with it a new category. It could be a vibe/genre, like Indie Pop, or an artist, like Beyonce. When you swipe down to this new category, a snippet of a song plays automatically alongside a looping piece of video content. Like it? Tap to listen to more, or maybe just add it to a playlist. Or would you rather pass? Well then swipe horizontally to the next rec, which is similar yet different than the last one you listened to. This horizontal swipe function similar to an Instagram story, allowing you to pop through a flurry of content strung along like beads on a string.
In this schema: the swipe down indicates a vibe shift. The swipe across though is a vibe perfection. Recommendations come to us in perfected, XY axis, precision.
Spotify has worked for years to make discovery and curation effortless for consumers. Since they learned in the early 2010s that news users were stressed out by choosing what music to listen to from the world’s history of recorded audio, they wanted to render themselves seamless. Discovery to these great tech-minds is a solvable problem. All it takes is some data and some nudges.
In a smart post that was partly responding to a past post of mine, writer Ock Sportello of the Never Hungover newsletter wrote, “Ask an avid TikTok user about their algorithm and ask yourself, when listening to their response, how differently they would have responded if you’d asked them to describe their personality.” It’s the kind of broad, simplistic statement I appreciate for being straightforwardly true. Our TikTok For You Page is a dialectic between algorithmic recommendations and us. The algorithm is a slab of marble, and with every like, comment, share, full-video-viewing, we shape our For You Page, if only slightly. In that sense, over time, the FYP does improve to more accurately reflect our tastes and desires.
This is not to say though that the TikTok algorithms reflects some deep insight into our vibe, though it may feel that way. I tend to watch a lot of TikTok content about reality TV, Dj’ing and queer-culture. These are not not my main interests, but I also contain multitudes and though my interests can be categorized, they cannot be encapsulated into a listicle. AI and algorithms can get pretty sophisticated and attenuated to the strange assemblage of my taste and vibe, but algorithms will always be chasing the shadow that is my own complexity. Because if the topics I engage with are new variables in a predictive algorithm, new key words in a data set, they will never fully get me, or any of us. I’m more than a series of topics. More than a codified set of affects. I pray I am at least.
Just as TikTok can reflect my interests but never my interiority with any sense of fine-grained detail, so too can Spotify’s recs never truly get me. I look at my Discovery Weekly for example and hear sounds and artists I enjoy, but like my TikTok topics, they are checking boxes. Weird-of-center experimental pop? Minimalist techno with a melodic, melodramatic bend? Nostalgic but contemporary hip hop? Indie acoustic folk with surprising electronic textures? Yup! And yawn!
Art, to use a big word, resonates not because it fulfills a need, but creates one. We find something that answers a question we hadn’t yet articulated. That’s what I turn to music for. I’m not “looking” for “chill indie acoustic rock with idm textures for some flare.” I’m looking to try and sense-make.
Algorithmic recommendations, or discovery through this narrow prism, will always be pivoting off a faint framework, codifying the ineffable into variable components. I’m sure a computer scientist or technologist can offer a hypothetical retort, but even if an algorithmic recommendation systems becomes exponentially better, algorithmic musical recs will always lack the context or necessary framework to be anything beyond checkboxes. There is not editorializing or contextualization in an algorithmic rec. You hear it. You like it. You save it.
Maybe that’s why music writing, fallow desert she may be, is a worthwhile project. If people are “discovering” music via algorithmic recs, and will only do so more in this swiping future, then it’s imperative that we create some space to contextualize the music that plugs into those predictive equations. (Shout outs No Bells which successfully contextualizes a scene in gorgeously rendered detail.) (And shout outs to Record Store where Amaya Lim provides thoughtful, contextualized playlists). I guess, for me, I’m trying to contextualize the algorithmic infrastructure that now defines so much of our listening.
But the thing about contextualization is it requires a little time and attention. To view a thing and notice it. Tough sell in a seamless and ever-giving universe. And I don't see Spotify’s TikTok knockoff as bringing us any closer to a more engaged, slower music listening community. All it suggests is a continued future of box checks, of careless swipes.
And I dont know the solution.
But for me, I am finding myself returning to obsession.
Because I just found a new favorite song. I’m going to play it til I hate it. At least, then, I’ll kind of, maybe, know something.
thanks for the shout tobias! and thanks for talking about how spotify keeps getting worse and worse... when do we jump ship ...