Drowning in Mediocre Data: On Kyle Chayka’s “Filterworld”

IN 1770, HUNGARIAN INVENTOR Wolfgang von Kempelen brought his latest creation to the royal court in Vienna, where he was eager to impress Empress Maria Theresa. The device consisted of a massive cabinet and a turbaned mannequin seated in front of a chessboard. It was called “The Mechanical Turk,” and it made short work of every challenger in the Viennese court. Von Kempelen eventually took the machine around the world, where it defeated the likes of Napoleon Bonaparte and Benjamin Franklin.

Of course, the whole thing turned out to be a hoax. Almost a century after the Turk beat its first opponent, it was revealed that the machine’s cabinet was hiding a presumably diminutive chess master who was surreptitiously choosing where the pieces would land next.

¤

How exactly did you end up here? I mean in the immediate sense: how did you, reader, end up on the Los Angeles Review of Books website, looking at this article, reading these words?

Most likely, you were led here by one of the great omnipotent algorithms, or at least by the shadow of one. The major gods—Twitter/X, Instagram, TikTok—anoint certain pieces of content seemingly at random, to the point that objective readings of causation and correlation are supplanted by a fractured internet mythology. Hopefully I’ve lit the right candles and said the right prayers for this piece to be swept into the current that separates the hits from the flotsam. Algorithms, take the wheel.

Investigating how the invisible hand of algorithmic influence has created an inscrutable, formulaic culture is the ambitious project of Kyle Chayka’s new book Filterworld: How Algorithms Flattened Culture. Chayka, a staff writer at The New Yorker whose beat can be roughly described as “the internet,” hit an algorithmic bonanza with one such successful essay. In 2016, he wrote an astute treatise for The Verge on the flattening of design taste into a minimalist international style he called “Airspace.” The critique put words to a trend we all had all been seeing through the filters of social media, where business owners catering to a moderately wealthy and internationally mobile demographic transformed hotel lobbies, coffee shops, and apartments into sanitized boxes that all look the same. (In Chayka’s estimation, much of this global aesthetic trend can be credited to the rise of Airbnb, hence the style’s name.)

It helps when an essay is a) about something observable to a wide swath of the population, and b) correct. “Welcome to Airspace” nailed both of those metrics. Everyone—or at least everyone who knows what The Verge is—had observed this interior design phenomenon. But to that point, no one had been able to analyze it in a way that was immediately digestible. Chayka scored bonus points for coining a neologism that, while not necessarily entrenched in the cultural lexicon, did provide a shorthand for something we could all see but not yet describe.

Where “Airspace” limited itself to critiquing the generic comfort of coffee shops and hotel lobbies, Filterworld widens the aperture to take in culture as a whole (though coffee shops still figure prominently). The topic is reminiscent of philosopher Timothy Morton’s “hyperobjects,” concepts so large, omnipresent, and influential that they become almost impossible to internalize. Algorithmic recommendation engines are the currents in which we are swept up as a society, choosing what we watch and listen to, sure, but also what we eventually find joy in. Filterworld starts from the beginning, rooting readers in the historical context of both mathematics and ghosts in machines.

Two threads run through Filterworld: one is the Mechanical Turk; the other is the technology of algorithms themselves. The mathematics at play are old. There is an algorithm written on a Babylonian tablet dated to sometime between 1800 and 1600 BCE. Apparently it can determine the square root of integers with a very high level of accuracy. (I haven’t tested it because that’s not how I make my money.) We didn’t get a name for algorithms until a couple thousand years later, when European scholars translated Kitab al-Jabr wa-l-Muqabala (“The Compendious Book on Calculation by Completion and Balancing”), the landmark work of Persian mathematician al-Khwarizmi, from Arabic to Latin. Those researchers also Latinized al-Khwarizmi’s name to “Algoritmi,” which was then used to describe mathematical processes. We also get the word “algebra” from the book’s shortened title (Al-Jabr).

The Mechanical Turk was retired in the 19th century, but Chayka considers the technological novelty a rough metaphor for the current state of things. “Algorithmic recommendations are the latest iteration of the Mechanical Turk: a series of human decisions that have been dressed up and automated as technological ones, at an inhuman scale and speed,” he argues early in Filterworld. Mechanical Turk is also the name of Amazon’s low-wage job service, where technology companies can hire armies of workers to do mundane tasks to give their platforms the illusion of automation. Silicon Valley hasn’t figured out how to scale irony yet, I suppose.

Chayka considers our modern Mechanical Turk and finds it less a parlor trick than an attempt at mass hypnosis. “Designed and maintained by the engineers of monopolistic tech companies, and running on data that we users continuously provide by logging in each day, the technology is both constructed by us and dominates us, manipulating our perceptions and attention,” he writes. “The algorithm always wins.”

¤

In the most basic sense, an algorithm is a set of instructions. The Euclidean algorithm from 300 BCE can determine the greatest common divisor of two whole numbers, which is useful if you ever need to find the greatest common divisor of two whole numbers. It was not until several thousand years later that we saw the emergence of algorithms designed specifically with computers in mind. Those were the work of Ada Lovelace, who somehow developed them decades before a computer even existed, but still saw their potential for automating human processes and expanding humanity’s ability to accomplish complex computations.

When you talk about Spotify or Instagram’s algorithm, what you’re really referring to are those platforms’ algorithmic recommendation engines. Those suggestions are using the data you create—songs you listen to, accounts you follow, videos you like—as inputs to feed you a predictive flow of additional content designed to keep you on the platform for as long as possible.

Chayka traces that history from social media’s nascent stages—when content was organized chronologically—to its present algorithmic state, where platforms are making educated guesses about what song, picture, or video is most likely to keep you from looking away. We create our own filter bubbles (“filterworlds,” if you will) by dint of our own consumption. Chayka writes that his own social media feed is dominated by meditative clips of people retiling their showers. Mine is mostly skateboard tricks and animals behaving badly.

That we dam our own endless stream of content seems to be widely understood by people of a certain age. Consider the fact that “the algorithm” has become ironic shorthand for the unwieldy currents of the internet: we talk about the algorithm the way we talk about phases of the moon. It feels like this great unknowable thing, a hyperobject that culture bends around. But Chayka points out that these equations are not manna—they’re deliberate corporate machinations designed to maximize profits for those corporations.

The power and ubiquity of these platforms has even led to something called “algorithmic anxiety,” which Chayka says “happens because there is a dramatically asymmetrical relationship between user and algorithm.” He continues:

Algorithmic anxiety places the burden of action on the individual, not the business—the user must change their behavior or risk disappearing. […] The effect goes back to the Mechanical Turk; we can’t always tell the difference between technology working and the illusion of technology working, but the perception may be just as impactful, in the end, as the reality.

Much of Filterworld is dedicated to exploring how those currents of algorithmic recommendations have created a morass of aesthetic and cultural sameness. The commentary about charmless interior design (more coffee shops!) and Instagram-optimized art feels a bit stale, as Chayka and other critics of internet culture have spent years analyzing those topics. So too can his personal anecdotes, such as the eeriness of an Amazon bookstore full of titles that sound like those you might buy your Reagan Democrat uncle for Christmas. The insidiousness of our own data being used against us by corporate marketing teams has been well-documented for over a decade. Remember when Target allegedly outed a teen girl’s pregnancy because they started sending her coupons for diapers and baby formula? That was in 2012!

That being said, Filterworld nearly vibrates when Chayka brings his background as an art critic and curator to the fore. He positions curators as a potential salve for our current cultural malaise, a sort of anti–Mechanical Turk that rejects computational sleights of hand in favor of deep, patient research. Chayka writes that algorithms are unable to mimic their ability to juxtapose and guide, leaving a hollowness to everything we consume on social media. “The slow process of curation works against the contextlessness, speed, and ephemerality that characterizes the Internet,” he argues. “Algorithmic feeds disrupt curated juxtapositions and make it that much harder to interpret the broad swath of culture, to figure out which themes join things together and which aspects set them apart.”

Chayka also commits a pair of separate subchapters to MoMA design and architecture curator Paola Antonelli and veteran DJ Paul Cavalconte, who have both built careers out of creating daisy chains of interpretations in different mediums. Their professions are no doubt at risk of being denigrated in favor of algorithmic alternatives—Chayka points out Spotify’s AI DJ as an especially depressing development—but I was hoping that he would locate the current threat in the longer timeline of capital’s influence over art. Power brokers have consistently tried to convince us of artists’ talent if it serves a specific purpose. (See: Milli Vanilli. The Monkees. Jeff Koons.) Commerce has been winning this game for much longer than Facebook’s algorithm has been around.

¤

As with all books written about the present, Filterworld has a blind spot for events in the immediate run-up to its publication. The emergence of generative AI platforms like OpenAI’s ChatGPT in late 2022 has accelerated our sprint towards culture’s inoffensive middle. (To his credit, Chayka has spent the better part of the last 12 months writing for The New Yorker about how AI is changing culture, so consider those pieces something of an addendum to the book.)

Generative AI programs are also complex sets of recommendation algorithms trained on massive datasets in order to provide something resembling creativity. The simplest way to think of them is as machines designed to be very good at guessing the order of things based on the libraries of information they’ve been trained on. Early last year, science fiction author Ted Chiang, who has a background in technical writing, wrote perhaps the clearest explanation of how platforms like ChatGPT work.

In many ways, those textual predictions are part of the current Chayka is trying to turn:

My argument for the kind of curation that Antonelli and others practice is that the act of putting one thing next to another is an incredibly important one and should be left to people with deep knowledge about or passion for the subject at hand—people who care about the significance of proximity.

The main issue with treating the output from programs like ChatGPT as sui generis is that it is computationally designed to create the most predictable string of words possible. It cannot be anything but a churned average.

Consider the salivating Hollywood executives who knew generative AI programs would be especially useful in creating infinite versions of basic screenplays. It’s not that they thought ChatGPT could give them the next Die Hard or Moonlight or Parasite; it’s that they knew the highest profit margin existed at creating something so dull as to appeal to exactly enough people to turn a profit. (Professional human writers could be brought in to augment or elevate these scripts, but they would not own any of the underlying intellectual property.) The executives wanted to bring the philosophy of scaled mediocrity a few hundred miles south from Silicon Valley and wrap it up in celluloid.

¤

In the final section of Filterworld, Chayka attempts to throttle his own engagement with recommendation algorithms by taking an extended break from Twitter, Instagram, TikTok, and Facebook. He is lured back because of professional obligations—Chayka’s journalism covers the internet, after all—and an understandable desire for a passive connection to his social circles. Still, the hiatus did some beneficial rewiring. “I was surprised to find that something had changed in my brain chemistry, even during my relatively short absence from social media,” he writes. “My pace of consumption had slowed, and I was much more deliberate in selecting what to read, listen to, or watch. When I returned, the feeds felt too fast and chaotic, too far from chronological.”

Chayka is not a luddite. Rather, he sees the recent development of algorithmic recommendations put forth by massive technology companies as a corruption of the internet’s initial promise and a cultural gravity well into which otherwise propulsive vectors have been sucked:

Even in the short time of their rise, algorithmic recommendations have warped everything from visual art to product design, songwriting, choreography, urbanism, food, and fashion. […] It must make them tap the Like or Share button, or prevent them from hitting Stop or Skip, anything that would interrupt the feed.

The twin pressures for creators to inspire engagement and avoid alienation have meant that so many cultural forms have become both more immediately enticing and more evanescent, leaving behind nothing but an atmosphere. This enforced ephemerality of feeling and impermanence of context has hollowed out contemporary culture, leaving it less experimental and powerful than it might have been otherwise, without these pressures.

In Chayka’s estimation, we have entered a cultural cul-de-sac. The internet is boring. Twitter is a hellscape. We’re all just content machines attempting to hit the algorithmic jackpot. It’s a grim portrait. But Chayka remains hopeful: he points to the niche communities and platforms—the “fair trade” music-streaming service IDAGIO; the cinephilic joys of the Criterion Channel—that are reminiscent of the internet many of us remember from the 1990s and early 2000s. They are slower, without the computational illusions of taste—in other words, anti-algorithmic.

Whether people can be convinced to exit their infinite scrolls is still an open question, though. Filterworld’s place in the growing bibliography of modern internet criticism sits somewhere between schematic and critique. Explaining how the machine works is no doubt valuable, but if you use the internet with any regularity, you understand the inscrutability of the algorithm and its impacts on what you consume. Knowing how you ended up somewhere is only important if you care about the journey.