Hawkins found the variety refreshing, and after two years, he left his job. Every few months, he let a computer pick the city he would live in, based on airfare, cost-of-living estimates, and his projected income as a freelance programmer. He tried listening to music picked randomly by Spotify, wearing clothes bought randomly on Amazon, growing out random styles of facial hair, and arranging phone calls with friends on randomly selected topics.
Hawkins even chose to get a random tattoo based on a random image search. His commitment to the experiment was steadfast and somewhat terrifying. “I was really worried it was going to be anime porn or something,” he said, “and I’d be stuck with it for the rest of my life.” But the computer’s selection was an abstract illustration of a parent and child. “I super lucked out,” he said. (Other outcomes of his experiments were, perhaps inevitably, less positive—like the time the computer told him to grow a soul patch.)
Hawkins was living out a sort of algorithmic jujitsu, using his own code to redirect recommendation engines toward the unexpected. “The nice thing about randomness is that it can give you something that is completely outside of what you would even imagine,” he told me. “And one place that computers can benefit us is that they have such a wider range of things that they can be aware of.”
Unlike Hawkins, I couldn’t fathom letting randomness rule my life. But when he put it like that, I felt a bit sad. In the greater scheme of things, so little computing power is harnessed to promote variety—and so much is channeled toward predictability.
Facebook’s algorithm didn’t seem to know what to make of my newfound penchant for randomness. Once, I joined a group for python fanatics, and Facebook recommended that I check out Reptile Connection. I joined Dinosaurs, and the site suggested Paleo World. At times, my quest felt like pushing together two magnets of the same charge; I sought something different, but was steered toward more of what I’d already seen.
Facebook, Instagram, and other personalized platforms simply aren’t built for what Deng, Abidin, Hawkins, and I were doing. Their business model—targeting advertisements based on users’ demonstrated preferences—depends on anticipating what will be relevant to people and using that information to entice them to buy products and services. Showcasing random interests wouldn’t serve that goal.
The web hasn’t always been structured like this. In the 1980s and ’90s, Zuckerman told me, before platforms like MySpace and Friendster began to map the internet around our preexisting social lives, networks were usually organized by topic, with no algorithm steering you to what it thought you’d like. This older internet, he noted, was homophilous as well, with a highly educated, demographically narrow user base. But it was typical, say, for three cat lovers to have a conversation across three different continents. People frequently had meaningful interactions with strangers.