The animating ethos of the internet is Paul’s lament in the Letter to the Romans: “What I do, I do not understand. For I do not do what I want, but I do what I hate.” The genius of the internet’s algorithmic feeds is the way they make choosing invisible, defusing the culpability of both the user and the Big Tech behemoth, offering just one more thing to look at, without the responsibility of a click.

For technology and culture writer Kyle Chayka, the author of Filterworld: How Algorithms Flattened Culture, the key question is how we can reintroduce friction and deliberate choice in the face of a culture of autoplay and absence. Increasingly, when people read or watch something on the internet, it’s not the result of the user’s active choice. Algorithmic feeds rely on a user’s previous actions to decide what to show him or her next — an endless buffet without the burden of filling your own plate.

When Facebook, Instagram, and the artist formerly known as Twitter shifted from chronological feeds to algorithmic ones, the change was framed as a response to users’ failure to choose. If you follow too many people, it’s impossible ever to be “caught up.” You always might be missing the thing you most want to see. The algorithm was intended to do the filtering that users couldn’t quite bear to do — you’d never have to explicitly unfollow someone, but the content you were least engaged by would eventually vanish from your feed, and you wouldn’t note its absence.

The algorithmic feed promised personal customization, but that curation would be invisible. The user shouldn’t do very much work under the hood, specifying what he or she would like to see less or more of — the website would learn and tune the feed to keep the user happy. Of course, “happy” was hard to measure, so, as a first approximation, it would mean the user stayed on the site, still scrolling or watching, actively if shallowly engaged.

Chayka was curious about invisible curation and aesthetic convergence, especially because he was beginning to notice how, wherever he went, it felt like his arrival was anticipated. Something of a digital nomad, Chayka found that all over the world, new hipster coffee shops were opening like tiny embassies — home turf for him, complete with cortados, reclaimed wood, and industrial lighting. He termed the aesthetic “AirSpace” and was unsettled by his own longing for the comfort of sameness: “I fit in in these non-places, . . . but identifying with a literally empty symbol was a strange exercise.”

In Chayka’s analysis, the end state of an algorithm-driven, competitive media environment is a yearning for oblivion. Burn out the dopamine circuits for long enough and “our natural reaction is to seek out culture that embraces nothingness, that blankets and soothes rather than challenges or surprises, as powerful artwork is meant to do.” This endgame was anticipated by C. S. Lewis more than 80 years ago, when in his Screwtape Letters he had a senior devil instruct a junior tempter to lead his “patient” into something that sounds a lot like doomscrolling:

You no longer need a good book, which he really likes, to keep him from his prayers or his work or his sleep; a column of advertisements in yesterday’s paper will do. . . . You can make him do nothing at all for long periods. You can keep him up late at night, not roistering, but staring at a dead fire in a cold room. All the healthy and outgoing activities which we want him to avoid can be inhibited and nothing given in return.

In the end, Lewis’s devil says, the patient will realize that “I spent most of my life in doing neither what I ought nor what I liked.” You don’t have to assume that the internet is operated by literal demons to be worried that its prevailing currents carry us far from our own best intentions. Sometimes, the algorithmic feed coarsens us by interspersing violent images from wars around the world with light jokes for ’90s kids. Other times, we ingrain bad habits when our clicks teach the feed to show us the dumbest, most infuriating thinkers on the other side of a political divide.

But, frequently, what I find myself resisting is the way the algorithm tempts me to let a lower good eclipse a higher good. Once you face the infinity of the global content stream, there’s enough low-quality but decent material out there to fill up your whole day. There’s nothing wrong with a Twitter feed of jokes about the cast of Frasier playing D&D, but the supply of quirky humor exceeds my attention budget. I need to aggressively and actively choose the best, not just passively consume the okay.

But the best operates contrary to the way the algorithms monitor our satisfaction. We pause, close the app, contemplate. Something worth our time provokes friction and demands silence. All of that looks like failure to an uncurious app. Algorithmic feeds are risk-averse. It is easier for them to keep recommending more of what you find satisfying enough than to take a chance with a risky suggestion. You liked that Tiny Desk Concert by a Broadway cast, so can I interest you in this one?

When I interact with an algorithmic feed, I seldom feel that the program is genuinely testing its hypotheses about me or trying to show me something it expects I don’t like, to double-check that its model is correct. I take on a little more of that responsibility myself, trying to get engaged by some works that don’t feel like natural fits, so I have the chance to stretch a little and surprise myself. (Sorry it didn’t work out, Jon Fosse; I hope the Nobel is a consolation.)

For readers and watchers, the algorithm is a source of suggestions; for creators, it’s a tricky-to-please switchboard operator. Chayka is curious about how the flattening nature of algorithmic feeds affects business and artists. He concludes that the context collapse of algorithm-driven feeds means that artists have little ability to present their art as they intend it. For members of the band Galaxie 500, speaking through the algorithm feels like losing their voice. As one of the members tells Chayka, it’s troubling that one of the songs they think sounds least like them is their most popular song on Spotify. The (very minimal) streaming royalties don’t seem like much compensation for making such a skewed first impression. For everyone but those with locked-down social-media accounts, the first impression you give is similarly framed through your most viral posts — which are likely to be your most inflammatory ones.

Relying on multiple untrustworthy algorithm partners to reach your audience means that it’s hard to communicate only along your chosen channel. If you’re an author, it would be ideal if you could also develop a visual aesthetic that could fill up an Instagram grid, or get a ring light and a breathy, intimate voice for confessional, five-hours-of-taping-to-appear-appropriately-spontaneous videos on TikTok. All businesses need to have some level of marketing savvy, but rather than placing an ad in a local paper, you’re sending out messages-in-a-bottle into a global maelstrom.

In response, services such as Substack (which I use) don’t just offer a smoother user interface or better revenue-sharing. What Substack is selling is the promise to fade into the background, to be more of a utility pipe than an active curator. Authors have ownership of their own email lists — no algorithm tweak leaves them trying to find a way to reappear in the feeds of fans who have already said they want to hear from them. It limits spontaneous-feeling discoverability and returns to word of mouth. You have to convince your readers to recommend an essay to a particular friend, rather than being retweeted into the world at large or boosted because you pleased the Ineffable Algorithm.

When it comes to curbing the sites that remain algorithm-driven, Chayka’s proposed solutions don’t seem to match the magnitude of the problem he’s defined. He’s excited about the idea of stripping Section 230 speech protections from online-media companies that rely on algorithmic feeds. Once you use weightings to decide which posts to show a user and in what order, you’re not a pipe, you’re a publisher, he argues. Removing the Section 230 protections from algorithmic feeds would make them so legally hazardous for their purveyors that they’d quickly be withdrawn, or, Chayka speculates, become populated with a limited range of white-listed, pre-screened content.

His other proposal is to extend the logic of the Children’s Television Act to the internet, so that all content providers would be obliged to show us something good along with whatever sludge we and the algorithm choose together. But the intent of the Children’s Television Act wasn’t to get adults to watch Mister Rogers (as much as they might have benefitted from it); it was to force television channels to help provide a public good.

These feeds aren’t breaking the law, but they are breaking the culture. The most plausible way to treat Big Tech is as a sin industry, like porn, cigarettes, and gambling. The state can increase the friction of consuming where it can, through age-gating, small fees, and encouraging public schools to ban phones, but we also need to make deliberate cultural shifts in our homes and social gatherings. If you wouldn’t put out ashtrays for a party by default, maybe consider setting up a box for everyone to check their phone when they arrive. Saying no to the thin gruel of the feeds is easier when you set the social table with something more robust and nourishing.