Filterworld: The invention of the feed
my new book in progress
Hello! It’s been too long since I’ve sent out a personal email, but I’m hoping this summer will bring more as I continue working on my upcoming book, Filterworld. The book will be about how the algorithmic feeds of digital platforms have flattened and homogenized culture. This newsletter has an original essay that I wrote explaining some of my thinking on that subject, and then further down (belated) highlights of my work lately, New Yorker column and otherwise.
The Invention of the Feed
Filterworld is about how new technologies — algorithmic recommendations, for example — have consequences far outside the bounds of what we usually think of as technological. This is often the case with new devices: The telephone didn’t just allow for voices to travel over great distances; it fundamentally changed human relationships, the ways we socialize, and how business could be conducted. The printing press didn’t just make books as objects mass-producible; it changed the way knowledge was contained and disseminated, allowing more people to access literature than ever. There are these before-and-after moments in history that only become clear decades or centuries down the line. My gut instinct is that the algorithmic feed, an automatically sorted set of content tailored to a specific user, is the same kind of invention. We’re living in a moment of flux, when we don’t fully understand the technology’s consequences and yet have been suddenly immersed in it, the way motorcars were once unthinkable and then unavoidable.
Yet we have to be careful with the word “algorithm.” It’s a generic term that really just means something like equation: a process that can be repeated over and over again to solve a particular problem. But the “algorithmic feed” is something more specific. In the earlier years of social media, feeds — the main columns of information on a website, to spell it out literally — were linear, the term for being sorted chronologically. Twitter, Facebook, and Instagram showed you whichever posts were put up most recently, and by scrolling down you went back in time. That logic was set, consistent, and comprehensible. The newest stuff went first. But things changed, particularly in the middle of the 2010s. The social-media feeds became more algorithmic. I’m less wary about that phrase, using algorithmic as an adjective. It means that algorithmic recommendations occupied more space on the feed. The items were not in strictly chronological order, and chronology itself was less of a factor in terms of what got placed toward the top of the feed.
After this shift, Facebook wasn’t necessarily showing you what happened in the past hour; it showed what its recommendation system determined would be of most interest to you. Thus an article link that a professional acquaintance posted just moments ago might be buried under a wedding announcement posted last week by a friend. Various factors — how often you interact with the friend, how many friends you have in common, the content of the post, and how much interaction it received so far — were automatically judged to indicate that you would rather see the announcement than the article. (Or, at least, the algorithm judged that seeing the announcement was more likely to keep you engaged with Facebook in the long term.)
In an algorithmic feed, more people see content that is already popular, as determined by the actions of other users. There’s a clear snowball effect: When content gets some engagement, it gets more engagement. In the past, this might have happened slower and more organically: a band sells out a concert in their hometown, for example, so more people hear them, which sparks some radio play, which gets them the attention of a big record label, which produces a national hit. In the all-encompassing structure of the feed, a piece of music gets put out, perhaps as the soundtrack to a TikTok video. Immediately, it gets a little or a lot of engagement. If it gets too little, then it dies, cut off from audiences. If it gets a lot, then it might reach more listeners, then even more, until a single clip could hit a million TikTok views overnight, launching a musician into viral fame.
Of course, that fame could fade just as quickly. Artists have to prepare themselves in advance to be picked up by the algorithmic feed and then take the most advantage of it, while at the same time risking getting alienated by their own exposure. We don’t choose what kinds of content the feed accelerates — we can’t opt out of a specific post getting algorithmic promotion. In my experience, it’s only my worst, most banal tweets that go viral and get reposted by Instagram aggregation accounts. Sure, lots of people see them, but it’s not really the kind of writing I would like to be known for.
Linear feeds were messy and inefficient. You just saw whatever happened to be there when you logged on. If you were bored or business-minded, you could scroll through the entire feed until you got back to the stuff that you had seen last time. Some apps had labels for this stopping point, alerting you that you’ve seen every tweet or every Instagram story. That no longer happens; there is now so much atomized content online that it has to be sorted or else it would be incomprehensible at its current scale and pace. Algorithmic recommendations populate infinite feeds of ads and posts of accounts you don’t actually follow. TikTok doesn’t require following and often ignores chronology entirely; you’re as likely to see a video from weeks ago as the same day, and dates aren’t clearly marked. Engagement is the only variable that matters. It’s like if the nightly news were treated the same way as the random clips from “America’s Funniest Home Videos.”
When everything has been sorted for you in advance by a machine, and there is an endless supply of future options in the feed, the consumer experience becomes passive as well as confusing. We are not able to develop coherent, deepening relationships with the ideas or narratives that we follow. Instead, the algorithmic feed delivers increasingly meaningless stimuli (served up only in order to stimulate) falling into the zero-one binary of invisible or ubiquitous. Surely there were monks in scriptoria who found the sudden flood of printed books similarly overwhelming, a cheapening of skill that they had held tightly for centuries. Theirs was an artisanal culture, in which knowledge was transmitted from one person to one copy at a time, by hand. Printing presses let anyone just put text on a page! It must have been shocking.
In the end, printed books didn’t degrade culture. But I don’t think I’m being unreasonable when I argue that our current encounter with digital technology is much, much worse. Rather than the invention of the book, maybe the algorithmic feed is more like the gas-powered, four-person automobile: a technology that let individuals go too fast and too far on their own and has ended up being quite bad for the planet, to the point that we have to stop ourselves from using them.
Did you like this essay? Subscribe to this newsletter for more updates on my book progress and thoughts about the process. Send any feedback by replying or emailing email@example.com (book encouragement is great!), and follow me on Twitter @chaykak.
— You may not know that am the co-founder of Dirt, a daily newsletter about the vanguard of digital culture and cool stuff we consume online. My co-founder Daisy Alioto and I raised a round of investment to turn Dirt into a real media company, working in the space of web3! You can read about it in Axios but also subscribe.
— I’m really happy with how my New Yorker column on digital culture has been going this year. Here are a few that I think stand out:
How the Internet Turned Us Into Content Machines: Review of two recent books about the Internet’s impact on our lives and selves.
The Online Spaces that Enable Mass Shooters: On the Buffalo gunman, Discord, and the problematic private areas of the Internet.
The Life and Death of the Original Micro Apartments: On the deconstruction of the Nakagin Capsule Tower, an iconic Metabolist building in Tokyo.
Have iPhone Cameras Become Too Smart?: The latest iPhones have cameras that turn photos into algorithmically optimized digital images.
Watching the World’s First TikTok War: On the broadcasting of the invasion of Ukraine over TikTok, turning it into digital content.
The Promise of DAOs: Decentralized autonomous organizations are the latest craze in crypto, but their actual uses are not clear yet.
How Tumblr Became Popular for Being Obsolete: The 2010s social network is gaining users because of its outdated features, like a non-algorithmic feed.