How did video games get so complex?
Let’s take, for example, the hit 2011 game The Elder Scrolls V: Skyrim. It’s a Tolkien-like role-playing game, where you play as an adventurer within a sprawling medieval world. You fight scores of foes, from mace-wielding giants to fire-breathing dragons, and face hundreds of quests. Those quests increase in difficulty, so you’re constantly “leveling up” your character. Leveling up, in turn, requires pondering which of the eighteen possible skills (such as archery, lock picking, and alchemy) you ought to be mastering, which type of armor and weapon to buy and wield (there are thousands of combinations), and which bonus powers—“perks”—to purchase with your experience points. (There are more than 240 different perks.)
Even longtime gamers were stunned by Skyrim’s byzantine folds. “I’ve poured over a hundred hours into the game so far and still have no idea how many more quests may be lurking out there,” a game reviewer for Forbes magazine marveled.
Of course, video games used to be drop-dead simple. The first hit arcade game—1972’s Pong, a digital version of Ping-Pong—required only a few lines of instruction: “Deposit quarter/Ball will serve automatically/Avoid missing ball for high score.” As Pong innovator Nolan Bushnell once noted, it was a game “simple enough for a drunk to play.” The other early arcade greats were similarly stripped down. In Space Invaders, you shot the aliens as they thudded slowly down the screen. In Pac-Man, you avoided the ghosts and ate the dots. In Asteroids, you shot asteroids and UFOs and avoided getting hit by anything. Teenagers around the world crammed into arcades to try to master these simple games.
Then something happened: The teenagers began sharing information.
If you were hanging out in the arcades in the 1980s—as I was, in Toronto as a teenager—you spent a lot of time not merely playing the games but talking about them. While one kid played, a cluster of others would gather around the machine. You’d share bits of information about how best to beat them, learning tricks from your peers. Together, you began to discover odd quirks in the machines’ behavior. For example, elite Pac-Man players could figure out the dot-eating pattern that would allow you play for hours on a single quarter. There was also an exciting flaw in the game Galaga—the “no fire” bug. On the first stage, you could shoot all the aliens except for the bottom-left one. Then you let that one attack you and reattack while you dodged it, for fifteen minutes. If you did this, it would stop stop firing, and once you killed it, none of the aliens on any later levels would fire a single shot, making it possible to rack up a monstrously high score. Many games had similar bugs, little loopholes in the programming that created weaknesses a clever player could exploit.
If you were playing alone, discovering those bugs might take you hundreds of hours. But you weren’t playing alone. Arcades were social networks, and players were avid traders of information.
As a result, those early games never stood a chance. They gave up their few mysteries in a matter of weeks. Collectively, the network of gamers was too smart.
Pretty soon, designers noticed this collective intelligence of gamers—and began to respond. They made the games intentionally more complicated, filled with “Easter eggs,” little secrets to be uncovered. The first famous one was programmed into the Atari game Adventure. The designer, Warren Robinett, placed a tiny dot in one dungeon; if you somehow discovered it and picked it up, it’d open a gateway to a secret room that spelled out Robinett’s name. Over at Nintendo, designer Shigeru Miyamoto began including hidden zones in Super Mario games, where players could scoop up dozens of golden coins.
Then the Internet came along in the 1990s, and the collaborative smarts of gamers exploded. Gamers used Usenet discussion boards to pool their knowledge, creating lists of strategies. They became astonishingly fast at sussing out arcana and documenting it. When Sega released Virtua Fighter 3 in 1996, the game included eleven characters, with hundreds of hidden fighting moves. Barely a few months after the game had been released, I could find comprehensive lists of every character’s moves online, painstakingly assembled by gamers around the world. (Want to get the character Wolf Hawkfield to do a “double punch-elbow-double-arm suplex throw”? Just use this button combination: “Punch, punch, tap-forward, punch, tap-down, tap-backward, punch, kick, guard.”)
If a game didn’t have this sort of complexity, gamers got bored and moved on. The fun in a game isn’t in having mastered it. It’s in the process of mastery—of figuring out the invisible dynamics behind how the thing works, revealing its secrets. By the early 2000s, hidden material in a game was no longer window dressing. It had been transformed into one of the central pleasures. Designers were producing worlds jammed with increasingly obtuse puzzles, storylines, distant areas, and unspoken secrets, because gamers would complain if there wasn’t enough to figure out. And because they were working together collaboratively, they could figure almost anything out. In a sense, the collective smarts of players produced a cognitive arms race—with designers forced to produce ever more immense and complex imaginary universes.
This is precisely why something as insanely convoluted as Skyrim can exist. Individual gamers don’t find such a game daunting, because they’re not playing it as individuals. Within hours of the game’s release in November 2011, fans set about documenting its every nuance. Soon there was a wiki with 15,789 pages, including dozens of walk-throughs—descriptions of quests—complete with maps and screenshots, and boards teeming with tips and strategies on managing your inventory. (“You can get an invincible companion dog to fight for you with another human companion at the same time by speaking to Lod at Falkreath to find his dog Barbas.”)
To be sure, not all gamers use these collective documents. It’s fun to solve problems on your own, so research shows most players use crowdsourced documents sparingly, such as when they’re stuck or pressed for time. But when it comes to the biggest, most sprawling games—like World of Warcraft—the number of players who dip into crowd wisdom is huge: According to one estimate in 2008, about 50 percent of English-speaking World of Warcraft players were using the game’s wiki every month. Top players form “guilds” that play together, cooperating to tackle the most difficult monsters.
This collective intelligence shows up even in groups of gamers as small as two. Three years ago, the designer and developer Matt T. Wood began working on the sequel to the hit game Portal. In the original Portal, gamers attempted to escape puzzling rooms by using a “portal gun” that would let them teleport from one part of a room to another. The game was praised for posing some very tricky mind-and-physics-bending challenges. The sequel, Portal 2, included a co-op mode in which two people, each wielding his own portal gun, cooperated as they tried to jointly escape the rooms. Wood and his co-op teammates created puzzles that they thought were suitably difficult for a pair of gamers to solve together. But when they tested it out in real life, the co-op players breezed through much faster than Wood had expected. How could they be so good? The two players would often keep an audio channel open, brainstorming and building off each other’s ideas—so “that puzzle that would have taken you fifteen minutes to solve solo only takes you two minutes to solve with someone else working with you,” as Wood tells me in an e-mail. Two heads weren’t just better than one; they were exponentially so. “We didn’t anticipate how much harder the players would allow us to make it,” Wood notes. He and his teammates went back to the drawing board, crafting rooms that were far more challenging.
In a sense, the video-game world was the first major industry to encounter the cognitive power of a highly connected audience. And what did it learn? That a vast community of networked people isn’t just smart—it’s restless and hungry for complex problems. How can we tap that resource? What types of problems is collective wisdom good at solving?
• • •
The reason millions of people collaborate on playing video games is that millions of people can now collaborate on anything. As Clay Shirky wrote in his book Here Comes Everybody, society has always had latent groups—collections of people all obsessed with the same thing and wishing they could work together on it. This is what the theory of multiples would predict, of course: If you’re fascinated by subject X, no matter how obscure and idiosyncratic, a thousand people are out there with the same fascination.
But for most of history, people couldn’t engage in mass collaboration. It was too expensive. To organize a widespread group around a task in the pre-Internet period, you needed a central office, staff devoted to coordinating efforts, expensive forms of long-distance communication (telegraphs, phone lines, trains), somebody to buy pencils and paper clips and to manage inventory. These are known as transaction costs, and they’re huge. But there was no way around them. As Shirky points out, following the analysis of economist Ronald Coase’s 1937 article “The Nature of the Firm,” you either paid the heavy costs of organizing or you didn’t organize at all and got nothing done.
And so for centuries, people collaborated massively only on tasks that would make enough money to afford those costs. You could work together globally at building and selling profitable cars (like the Ford Motor Company) or running a world religion (like the Catholic Church), or even running a big nonprofit that could solicit mass donations (like UNICEF). Those organizations solved large, expensive, well-known problems—making cars, offering religion, helping the poor—that could generate serious cash flow. But what if a problem were smaller, more niche? Like how to find your way around a complicated video game? Well, there was basically no way to organize around it. Latent groups stayed latent.
Until the Internet came along. Now that self-organization online is basically free, those latent groups have burst into view. When a dozen friends spread across a city use a Facebook thread and a cute little voting app to pick which film they’ll see on Friday night—“vote for your favorite!”—they are engaging in the same collective decision making that was previously available only to well-funded organizations. This, again, is basic behavioral economics: If you make it easier for people to do something, they’ll do more of it. Finding your way around Skyrim or resolving conundrums like “Which movie are we seeing tonight?” are problems that traditionally couldn’t afford Ronald Coase–style transactional costs—they fell “under the Coasean floor,” as Shirky puts it. But things have decisively changed. “Because we can now reach beneath the Coasean floor,” he writes, “we can have groups that operate with a birthday party’s informality and a multinational’s scope. . . . Now that group-forming has gone from hard to ridiculously easy, we are seeing an explosion of experiments with new groups and new kinds of groups.”
The stuff that lives beneath the Coasean floor tends to be incredibly weird. I say that as a compliment. In a world where people think publicly and harness multiples to find like-minded souls, we can find other people worldwide who share our marginal interests, the sub rosa hobbies that we long nourished but couldn’t find anyone geographically nearby to share. The result is a flood of amateur collaboration. I’m using “amateur” in its original sense, meaning not “done poorly” but “done for love instead of money.” One of the reason some cultural elitists—political pundits, novelists, intellectuals—tend to be so unsettled by the Internet is that it has revealed how oceanically broad are the interests of the public in general. Before the Internet, with no way of observing the obsessions of the masses, it was a lot easier to pretend that these obsessions simply didn’t exist; that the nation was “united” around caring about the same small number of movies, weekly magazines, novels, political issues, or personalities. This was probably always a self-flattering illusion for the folks who ran things. The Internet destroyed it. When you gaze with wild surmise upon the Pacific of strangeness online, you confront the astonishing diversity of human passion.
• • •
Take, for example, fan fiction. This is the art of writing stories based on one’s favorite cultural products, such as novels, movies, or TV shows. It’s an old practice, dating back at least to the early twentieth century, but in the 1970s, new variants began to emerge, spinning off from shows like Star Trek. One vibrant subset was slash fiction, written by Star Trek fans intrigued by the idea of a homoerotic relationship between Kirk and Spock—a wild, Heathcliffean hothead paired with a cool, distant partner. They started writing stories in which Kirk and Spock had an actual relationship, including plenty of steamy sex scenes, sometimes circulating the stories in photocopied zines. Pretty soon fans, very often women, were penning similarly riotous tales of other male couples from mainstream TV, including Starsky/Hutch pairings or Blake/Avon pairings, from the show Blake’s 7. (Hence “slash” fiction, for the “/”.) Other forms of fan fiction produced heterosexual couplings: The X-Files produced an avalanche of stories exploring the romantic life of Mulder and Scully, and Harry Potter produced Hermione/Harry tales. And plenty of fan fiction today has no sex at all; some fans just enjoy writing and love tinkering with their favorite fictional universes. The literary form gained some mainstream prominence in 2012 with the publication of Fifty Shades of Grey, a book whose author originally developed the story as Twilight fan fiction.
It’s a subculture rife with multiples: finally all those folks in Iowa who’d been secretly penning tales based on CHiPS could discover those in Germany doing the same thing. These fans have been at the forefront of using digital tools to connect with one another. They became early expert users of bookmarking sites, putting tags on their stories—codes highlighting specific qualities—so readers could find exactly what they wanted. “If you wanted to read a 3000 word fic where Picard forces Gandalf into sexual bondage, and it seems unconsensual but secretly both want it, and it’s R-explicit but not NC-17 explicit, all you had to do was search along the appropriate combination of tags,” as Maciej Ceglowski, the founder of the bookmarking site Pinboard, wrote in his blog.
In 2011, Ceglowski had noticed a surge of fan fiction folks using his service. Since this was a profitable new base of customers, he figured he should find out if they had any requests. Should he add new features to Pinboard? On September 28 at 4:19 p.m. Pacific time, he tweeted an open-ended request: “Fanfic people, can you draft a list of your ‘must have’ features for me to look at, maybe as a Google doc? I’ll implement what I can for you.”
What resulted was an astonishing display of collaborative thinking. The fan fiction community instantly began retweeting Ceglowski’s invitation. Within minutes, dozens from around the world had set up an open Google Docs document and began writing a wish list of features. That list quickly sprawled to thousands of words and became unreadably complicated, so the fans crowded into the “chat” area for the document and began editing it. One fan sorted the most common requests; another began carefully formatting the document to make it easier to read; others corrected typos. Within two days, they’d written a meticulously crafted sixteen-thousand-word “design document” for Ceglowski—mapping out how Pinboard could evolve, including samples of code he could use. He was astonished. “These people,” he wrote later, “do not waste time.”
The fans were almost as surprised as Ceglowski at how quickly and smartly they worked. “No one was in control. People were going, ‘Where did all these people come from?’” as Priscilla Del Cima, a twenty-five-year-old graduate student from Rio de Janeiro who worked on the document for forty-eight hours, pausing only for brief snatches of sleep, tells me. There were so many people crowded into the chat area that it crashed (“Google Docs can’t handle more than fifty people chatting at a time”). She’s not sure precisely how many fans were involved overall—hundreds, she guesses—but the group was diverse enough that it tapped into a broad range of skills. “There were some very tech-savvy people on the chat, so when we started asking for something really hard they could say, ‘Okay, that’s technologically feasible and that’s not.’”
This breadth of participation is key to what author James Surowiecki dubbed “the wisdom of crowds.” Crowd wisdom as a scientific phenomenon was first explored in 1906 when the British scientist Francis Galton visited a county fair and observed a contest to guess the weight of an ox. About eight hundred fair attendees put in guesses. Galton expected that compared to the guess of an expert judge of oxen, the crowd of attendees would be far off the mark. But when Galton averaged out the crowd’s guesses, the result was 1,197 pounds—just one pound less than the actual weight. “In other words,” Surowiecki writes, “the crowd’s judgment was essentially perfect.”
How could that be? It’s because, Surowiecki argues, each member of a decent-sized crowd of people possesses some incomplete part of the picture in her head. If you have a mechanism to assemble the various parts, you can wind up with a remarkably complete picture. The reason we never knew this was that for centuries, we had few mechanisms for assembling people’s collective judgments. The ones that existed were large and professionally run, like polling firms or governments running elections. Now that ordinary people have mechanisms for aggregating knowledge, ever more complete pictures are swimming into view. The scads of Pinboard memo contributors each had one idea for improving the service; the thousands of Skyrim players each observed one small, stray fact about the game.
We see these mass think-ins happen frequently in response to cultural challenges, because people share so many cultural passions today. But group brainpower just as easily coalesces around difficult political problems, too.
Take the remarkable case of Tahrir Supplies in Egypt. In the fall of 2011, the Egyptian activists who had driven Hosni Mubarak from power faced a new danger: the Egyptian military. The military had assumed control and wasn’t showing any signs of ceding to democratic control. So the protesters again massed in Cairo’s Tahrir Square, where their movement had originally broken out. In November, on Mohamed Mahmoud Street just east of Tahrir Square, security forces moved in, unleashing torrents of tear gas and heavy beatings, even firing upon the protestors with live ammunition. The activists set up ten makeshift tent hospitals to treat the wounded.
Sitting in his apartment far away in Dubai, a twenty-three-year-old Egyptian named Ahmed Abulhassan was watching the fight via online video clips and Twitter. He had recently left Egypt after graduating from university with a degree in pharmacy biotechnology, but he knew dozens of friends who were back home being attacked by the military. “I was following the news and getting infuriated by it,” he tells me.
But Abulhassan also noticed that the tent hospitals had a coordination problem. Doctors and volunteers with supplies would show up at one tent, but they’d actually be needed at a tent several blocks away. In the confusion, it was hard to know where to go. The activists did not have good “situational awareness,” and Abulhassan realized he could use online tools to provide it.
He started a Twitter account called @tahrirsupplies, devoted solely to reporting which tent hospitals needed supplies. To get the word out, he begged various Egyptian celebrities and popular Twitter activists with hundreds of thousands of followers to retweet him.
It worked. Within a day, @tahrirsupplies had ten thousand followers, and activists were flooding Abulhassan with hundreds of texts and phone calls and tweets. He would learn which hospitals had which needs, then turn around and tweet that information. Doctors would know where to go, while Egyptian citizens would dash to their nearest pharmacy, buy supplies, and rush them over.
Abulhassan created, in essence, a highly collaborative, on-the-fly aid organization. It was like Médecins Sans Frontières, except without any head office or staff at all. As the battle raged for six days, requests got bigger. Three Egyptian women in their early twenties offered to help in the coordination. None were actually physically present; one lived in England, and two lived in Cairo but had parents who forbade them from joining the street action. “They said, ‘Can we help?’” Abulhassan says, “and I said, ‘Absolutely! This is not a business. I’m not owning it.’” Over the four days, the group slept in shifts, to make sure someone was awake to tweet requests.
Soon the requests were so numerous, they set up a Google Docs speadsheet to track the array of necessary goods, like Hemostop to seal wounds and Ventolin spray to heal lungs scarred by tear gas. The team even received requests for electrical generators and “an eye surgery machine for anesthetizing eyes.” (Sure enough, wealthy Cairo residents bought and delivered those, too.) When pharmacists visited the hospitals to take an inventory, they estimated Tahrir Supplies had helped coordinate the delivery of one million U.S. dollars’ worth of medical supplies.
Abulhassan never saw a penny, of course. Technically, his organization didn’t exist. It was just a vehicle for helping people to help each other.
After the battle died down, Tahrir Supplies kept going, using its network to help out with other issues. It connected patients with rare blood needs—“we find people who are O negative and they both contact each other,” Abulhassan says—and created medical caravans for poor villages. And when conflict broke out in Tahrir Square over the next year, it set about organizing aid there again.
In early 2012, Abulhassan found his first job after graduating. During the interviews, they asked him the standard interview question: How did you solve a problem in an innovative way? He laughed. For the first time, he had a good answer.
• • •
Collaboration and collective thinking don’t happen by magic. In fact, they follow several rules, which you can see in action in these success stories.
First, collective thinking requires a focused problem to solve. With Tahrir Supplies, Abulhassan was trying to answer a clear question: Which tent hospitals in Cairo need help, and what do they need? This made it easier for a mass of people to contribute information. Indeed, he specifically avoided using his Twitter stream for anything but straightforward, verifiable factual information about the status of each hospital. (He didn’t post thoughts about politics, even though, with a huge audience, he was tempted. “People were asking, ‘What’s your message?’ But we were just trying to do one job.”) Similarly, the contributors to the Skyrim wiki are united around a desire to produce a list of facts that will help players navigate a complex game.
Without a clear goal, collective thinkers can zoom wildly off base. Consider the spectacular collapse of the Los Angeles Times’s “wikitorial.” Inspired by Wikipedia, the paper published a one-thousand-word editorial about the Iraq War as a wiki, then encouraged readers to edit it. “Do you see fatuous reasoning, a selective reading of the facts, a lack of poetry?” the Times’s editors wrote. “Well, what are you going to do about it? . . . Rewrite the editorial yourself.” Within a day, the project began to unravel. Several hundred readers tweaked the text, but they couldn’t agree on what direction to take it. Several made the tone more harsh: “The Bush administration should be publicly charged and tried for war crimes and crimes against humanity,” one editor amended, and another simply wrote, “Fuck USA.”
The problem with the wikitorial was that the goal had no obvious end point. A large group can argue about a set of facts and come to a reasonable consensus; Wikipedia does this every day. But a strongly worded opinion—the core of an op-ed—is not subject to consensus. This is why collective thinking online also tends to fail when it attempts an aesthetic creation. The Web designer Kevan Davis set up an online experiment to group-think pictures, allowing anyone to vote on the color of a randomly located individual pixel. But when the group tried to draw relatively simple figures—a castle, an apple, a human head, a cat—it produced undifferentiated blobs. Nobody could agree on what the subject ought to look like. Which is the point: Art is usually the product of a single independent vision. As most corporations discover to their dismay, groups can suck creativity out of projects because they tamp down the most original, idiosyncratic parts of each individual’s vision.
Collective puzzle solving also requires a mix of contributors. Specifically, it needs to have really big central contributors—and then a ton of people making microcontributions.
It needs central people because, despite often being called leaderless, these projects are rarely entirely so. In reality, they almost always rely on a core group of contributors—the folks who frame the problem and get the project rolling. With Tahrir Supplies, it was Abulhassan; with the Pinboard project, while no one appointed themselves the leader, there were about a dozen very committed fans (Del Cima was one) who did crucial work like formatting the document and weeding out duplicate ideas. These hard jobs require enthusiasm and a time commitment, so they have a higher barrier to entry.
But they’re not enough. Really successful collective thinking also requires dilettantes—people who offer a single small bit of help, like doing one edit or adding a fact or photo. Though each microcontribution is a small grain of sand, when you get thousands or millions you quickly build a beach. Microcontributions also diversify the knowledge pool. If anyone who’s interested can briefly help out, almost everyone does, and soon the project is tapping into broad expertise: “The small contributions help the collaboration rapidly explore a much broader range of ideas than would otherwise be the case,” as the author Michael Nielsen notes in Reinventing Discovery, an investigation of crowdsourced science.
Tahrir Supplies leveraged microcontributions brilliantly. Because Abulhassan made it so easy for activists to contribute news—all they needed to do was text, tweet at, or e-mail the central team—thousands did, and together they amassed a more complete picture of the situation than the central team could have achieved on its own. We see this pattern at Wikipedia, too. Because it’s so easy to add to a Wikipedia article—hit “edit” and boom, you’re a contributor—the scale of microcontributions is vast, well into the hundreds of millions. Indeed, the single most common edit on Wikipedia is someone changing a word or phrase: a teensy contribution, truly a grain of sand. Yet, like Tahrir Supplies, Wikipedia also relies on a small core of heavily involved contributors. Indeed, if you look at the number of really active contributors, the ones who make more than a hundred edits a month, there are not quite thirty-five hundred. If you drill down to the really committed folks—the administrators who deal with vandalism, among other things—there are only six or seven hundred active ones. Wikipedia contributions form a classic long-tail distribution, with a small passionate bunch at one end, followed by a line of grain-of-sand contributors that fades off over the horizon.
These hard-core and lightweight contributors form a symbiotic whole. Without the microcontributors, Wikipedia wouldn’t grow as quickly, and it would have a much more narrow knowledge base. (And Wikipedia’s base of microcontributors still needs to become even more diverse, frankly. Because Wikipedia’s hard-core contributors tend to be mostly techie white men; it has terrific coverage of computer science and physics—but big holes in other areas. “I go to find an article about Swedish feminism because I’m reading Stieg Larsson, and it doesn’t exist,” as former Wikimedia Foundation executive director Sue Gardner once told me; she’s been trying to recruit more women and non-Western contributors.)
This blend of microcontributors and heavy contributors works well for gathering and organizing data. But it even works with “insight” problems—the type that require an aha breakthrough. Consider again the world of chess, where such epiphanies are critical. In 1999, Garry Kasparov engaged in yet another fascinating experiment by playing a game against an online collective. Billed as “Kasparov versus the World,” it allowed anyone interested in the game to visit a Web site where they could suggest a move and vote on the best one to play against Kasparov. More than fifty thousand people participated from more than seventy-five countries, with an average of five thousand people voting on each move. The group contained a few heavy contributors, most notably the fifteen-year-old Irina Krush. A rising star in the chess world, Krush had recently become the U.S. women’s chess champion. At move 10, Krush suggested a play—which the collective adopted—so powerful that Kasparov called it “an important contribution to chess.” As Nielsen points out in Reinventing Discovery, Krush also helped coordinate the mob: Together with her management team, she created an online “analysis tree” that listed possible gambits, so the group could stay on the same page as they voted.
But the microcontributions and cognitive breadth were crucial. Having one really smart player, or even a small handful, wasn’t enough to fight Kasparov. Krush said one of her three favorite moves the team played was number 26, which came not from her but from Yaaqov Vaingorten, “a reasonably serious but not elite junior player.” Thousands of other players offered tiny bits of analysis that shaped successful plays. In the end, Kasparov still triumphed, but it took him sixty-two moves and several nail-biting reversals of fortune. Collectively, the group played far more skillfully than Krush or any of the group’s individual players could. Kasparov called it “the greatest game in the history of chess,” adding, “The sheer number of ideas, the complexity, and the contribution it has made to chess make it the most important game ever played.”
There’s one final, and very subtle, part of smart collective thinking: culture. It turns out that the type of people in the group and the way they interact spell the difference between success and failure.
Wikipedia is perhaps the most famous collaborative project. The real secret to its success, though, isn’t merely its millions of volunteers. It is, as communications professor Joseph Reagle dubs it, the culture of “good faith collaboration” that Wikipedia cofounder Jimmy Wales labored to put in place—a commitment to Quaker-level civility. Not long after launching Wikipedia, Wales penned an open letter to all potential contributors (by which he meant the entire planet), arguing that the project would only work if the contributors struggled constantly to remain polite to one another. “Mutual respect and a reasonable approach to disagreement are essential . . . on this incredible ridiculous crazy fun project to change the world,” Wales wrote. This attitude was later codified by the users themselves as one of Wikipedia’s Five Pillars of self-governance:
Respect your fellow Wikipedians, even when you disagree. Apply Wikipedia etiquette, and avoid personal attacks. Seek consensus, avoid edit wars, and never disrupt Wikipedia to illustrate a point. Act in good faith, and assume good faith on the part of others. Be open and welcoming.
Another of Wikipedia’s Five Pillars addresses the site’s “neutral point of view”: “We avoid advocacy and we characterize information and issues rather than debate them.” Indeed, faced with a controversial subject about which she feels strongly, a Wikipedia contributor ought to work extra hard to carefully describe views she finds repellent. Since Wikipedia contributors regularly disagree about facts—ranging from hot-button issues like “when life begins” on the abortion page to whether Star Wars media ever actually identify Yoda’s home planet—the only thing keeping articles from being endlessly rewritten by warring factions is for the factions to stop warring. That can’t be done by software; it takes culture. As Reagle points out, Wales spent countless hours illustrating his own principles by politely urging combatants to be, well, polite. To defuse the sense of urgency that often makes arguments more bitter, Wales would point out that actually there’s no rush at all in working on Wikipedia, and in fact “there is plenty of time to stop and ask questions.” The upshot, Reagle notes, is that the interactions among Wikipedians often continue at prodigious length. They go on seemingly forever, which is a deficit (they can be mind numbing) and a delight (they permit and encourage the sort of crazy rathole exploration that leads to productive thinking). They are “frequently exasperating, often humorous, and occasionally profound,” as Reagle writes. Even white supremacists who’ve tried to edit Wikipedia pages—surely among the most adversarial, antagonistic contributors one could imagine—have absorbed this culture and “reminded themselves they need to be cordial on Wikipedia,” as Reagle has found.
• • •
To be really smart, though, an online group needs to obey one final rule—and a rather counterintuitive one. The members can’t have too much contact with one another. To work best, the members of a collective group ought to be able to think and work independently.
This rule came to light in 1958, when social scientists tested different techniques of brainstorming. They posed a thought-provoking question: If humans had an extra thumb on each hand, what benefits and problems would emerge? Then they had two different types of groups brainstorm answers. In one group, the members worked face-to-face; in the other group, the members each worked independently, then pooled their answers at the end. You might expect the people working face-to-face to be more productive, but that wasn’t the case. The team with independently working members produced almost twice as many ideas. Other studies confirmed these results. Traditional brainstorming simply doesn’t work as well as thinking alone, then pooling results.
That’s because, the scientists found, groups that have direct contact suffer from two problems. The big one is blocking—a great idea pops into your head, but by the time the group calls on you, you’ve forgotten it. The other is social dampening: outspoken, extroverted members wind up dominating, and their ideas get adopted by others, even if they’re not very good ones. Introverted members don’t speak up. In contrast, when group members work physically separately from one another—in what researchers call “virtual groups”—it avoids this problem because everyone can generate ideas without being cognitively overshadowed or blocked. This is one of the counterintuitive secrets behind online collaborations. They inherently fit the model of people working together intimately but remotely.
If we’re too readily swayed by the views of people when we’re in a room with them, is it possible to be similarly led astray by the views of people online? Apparently so. A 2011 study took several virtual groups and asked them to estimate obscure facts about Switzerland, such as its population density and crime rate. At first, the group members had no exposure to each other, and—like the folks guessing the weight of the ox—the group was smart. The average of their guesses was close to the facts.
Then the scientists changed the experiment. As the group members were cogitating alone in their cubicles, the scientists gave them more information: Each member could see the others’ guesses and the average of the group and then they could revise their guess. What happened? The wisdom of the group broke down. Members began to drift toward each other, influenced by the others’ guesses. This caused a bad feedback loop, because as each individual member revised his own guesses, the overall average guess grew less accurate. By being exposed to one another’s individual thinking too intimately, the group’s collective wisdom declined. It got stupider. The members began to suffer from precisely the same problem that can afflict face-to-face groups.
This problem, of course, crops up all the time online—the “rich get richer” cycle of popularity. When a newspaper puts a list of its most-e-mailed stories on its front page, it represents the collective judgment of its readers. But it’s also biasing incoming readers. The crowd knows too much about what the crowd thinks, and its wisdom vanishes in an Ouroboran gulp of its own tail.
Now, obviously we’re not completely robotic lemmings, otherwise the top-ten lists would never change. Even when exposed to one another’s views, we still hold our own counsel. But this question—“How much should an online crowd know about itself?”—turns out to be among the biggest design challenges for anyone trying to harness collective thinking. The trick is to encourage people to join in but also to think for themselves.
• • •
Historically, one way we measure intelligence is by evaluating one’s ability to solve problems. But in collective thinking, a new proposition emerges that flips this logic on its head. The key skill here is designing problems—designing them in a way that lets many people pitch in to solve it.
This is the way David Baker and his team cracked a ten-year-old biological puzzle. Baker is a biochemistry professor at the University of Washington who studies proteins, the devilishly complicated molecules that make up the human body. A protein is an incredibly long string of atoms folded into a tight ball—think of a long piece of yarn rolled tightly into a ball around itself. There are billions of possible ways to fold up a protein into a ball, each making the protein behave in a slightly different fashion. If you can figure out how the protein folds, you can understand how it works, and this is crucial in both understanding diseases and designing drugs to combat them.
For years, Baker had been studying protein folding with brute-force computer power. He’d create a computer model of a long protein string, then have his computer semirandomly fold it, hoping to hit upon a revealing result. The process was glacially slow; searching randomly is not very efficient. To amass more computing power, Baker created Rosetta@Home, a distributed computing application. You could download the program and run it as a screen saver; if your computer was sitting idle, it would start randomly folding proteins, effectively contributing free computer power to Baker’s academic research. Thousands of people downloaded Rosetta@Home, eager to help out the cause of science. It was also quite beautiful to watch, because you could see a graphic of the protein being folded on-screen.
Then Baker started getting e-mails from users. They’d watch the protein being folded and realize, Hey, I could do this better than the computer. The computer was dumbly trying random combinations. Humans could spy better ones. They wanted to try doing it themselves.
A collaborative-thinking project was born. Baker joined forces with some programmers at his university and created Fold.it, a program that let people fold a protein themselves. Users employed the mouse and buttons to tweak, jiggle, and twist proteins into increasingly elegant balls. Fold.it had a fun, gamelike interface: If you managed to produce an efficient fold, it would generate a high score. Pretty soon there were two hundred thousand players collaborating. They set up a wiki to discuss their favorite strategies, cramming into the Fold.it chat rooms to share tips. Some formed teams, with names like Void Crushers or Contenders, to achieve high scores. “Encouraging discussion and questions, all are free to express themselves,” as the Contenders explain on their Web site. “We play our soloist games our own way; but if someone finds sudden success, it’s posted for the benefit of the group, detailing what was done to get there.” In effect, Baker had designed a problem that perfectly leveraged collective thinking. It had a specific goal, brought in a diverse range of expertise, encouraged microcontributions, and let people think quietly on their own even while collaborating with others.
Soon, it led to scientific breakthroughs. In 2010, Baker decided to give the Fold.it community a particularly tricky challenge: folding the M-PMV virus, which causes AIDS in monkeys. Biologists had been trying for a decade to figure out how M-PMV folds but had arrived at only an incomplete solution. The Fold.it players quickly began improving on that solution, with one finding a cleverer fold, which another one would improve upon, and so on.
In only three weeks, the amateurs of the Void Crushers and Contenders teams had jointly solved a problem that had bedeviled professional biologists for ten years. Baker published the results in Nature Structural & Molecular Biology in an essay titled “Crystal structure of a monomeric retroviral protease solved by protein folding game players.”
Baker is continuing to use the Fold.it community to solve hard folding problems; in fact, he has published three more papers based on their successes. In one case, they came up with a “crazy” way to fold a binding enzyme that Baker and his colleagues had never considered. In another paper, Baker studied the strategies players were using to fold proteins and found they had independently hit upon techniques that biologists had been using in private but hadn’t published yet. How had the crowd been so innovative? Again, by being very open and sharing their best work and posting their favorite techniques so others could improve them. “The best ones would just go viral,” Baker tells me. Baker’s success wasn’t in finding a solution. It was in designing a good problem—a game system that neatly channels the capabilities of the group.
Motivation matters. People eagerly pitch in on projects they’re interested in, which is exactly why so many ad hoc collaborations erupt around amateur passions, like hobbies or pop culture. But as the Fold.it guys found, science projects work well, too, because people enjoy feeling they’re contributing to global knowledge. Many other scientists have joined Baker in crafting successful group thinking projects, such as the Galaxy Zoo, founded by the Radcliffe Observatory in Oxford, England, which puts a deluge of space imagery online and lets everyday astrophiles classify the shapes of galaxies; like Fold.it, it quickly evolved a community of contributors. Politics, too, is an area where many people are motivated to help, which is what drives the success of the Ushahidi maps or government 2.0 projects, where citizens compile information to improve their communities.
But can you make money off collective smarts? Can they help corporations work more intelligently?
It’s harder than you’d think. Compared to the many public-minded projects, few corporations have been able to harness huge, public groups of collective thinkers. Motivation is a problem: Few people think profitable companies deserve their free work. At best, companies have been able to deploy fairly simple polling- and-voting group thinking projects, often to tap in to what their customers want; clothing firms like Threadless let their users vote on user-submitted designs. Others have solved the motivational problem by offering substantial prizes. Netflix, for example, offered a one-million-dollar prize for whoever could improve its movie-recommendation algorithm by 10 percent. But while prizes motivate hard work, they can inhibit sharing. When people are competing for a big prize, they’re often not willing to talk about their smartest breakthrough ideas for fear that a rival will steal their work. (Indeed, as teams got closer to winning the Netflix prize, they became increasingly secretive.)
Other corporations have solved the problems of motivation and secrecy by turning inward and creating internal “decision markets” where employees can pose ideas and vote on the best ones. At one software firm, Rite-Solutions, an employee pitched a new product—a 3-D environment designed to help military clients handle emergencies—and after it got heavily upvoted, the firm decided to build it, whereupon it became the source of 30 percent of their annual sales. Internal markets can be extremely valuable, because they keep trade secrets secret. But they dramatically shrink the pool of people who help solve problems. As Tahrir Supplies discovered, when it comes to microcontributions, scale matters. If, as Sun Microsystems cofounder Bill Joy reportedly liked to say, the smartest people are outside the room, collective problem solving requires being as public as possible. That’s harder with for-profit challenges.
The most successful corporate think-ins emerge when consumers perceive an overlap between their self-interest and the self-interest of the corporation. Some newspapers have tapped crowd smarts to assist with investigative journalism, because enough people feel these investigations are a public good. If designed well, they’re also fun. When the UK government was forced to release a trove of expense receipts for British politicians, the unsorted pile was too big for any single journalist to sift through—so the Guardian created a gamelike online tool that let anyone comb through the stack and flag dodgy-looking expenses. More than twenty thousand citizens analyzed a stunning 170,000 documents in four days, and the Guardian published a story listing some of the crowd’s most egregious discoveries (such as an MP who’d charged £225—$441—for a duvet.) Several political sites have used their large passionate audiences to comb through similar governmental data dumps. And to figure out which neighborhoods charge the most for groceries, fans of the Brian Lehrer radio show in New York sent in the price of lettuce, beer, and milk in their local markets, which produced a witty map of the city.
Google, which relies on analyzing links, has essentially built its search empire on collective knowledge. Every time someone posts a link to a Web site, they’re giving Google information to analyze; each link is a tiny vote for the site’s relevance. Several other Google projects have leveraged different types of collective effort. When I visited the offices of Google Earth, its product manager, Peter Birch, booted up the software and zoomed in to Red Square in Moscow. As we approached street level, I could see hundreds of buildings appear, perfectly modeled in 3-D, including gorgeously rendered versions of St. Basil’s Cathedral, with its colorful, bulb-topped towers. Google didn’t design those buildings; fans of 3-D modeling did. Google simply made it easy to contribute, releasing free Building Maker software and an online tool for submitting your building for inclusion in Google Earth. If it’s accepted, Google includes your user name in the model, so people can know who made it and see all your other buildings, too. “Now we have an amazing amount of buildings all over the world,” Birch told me, hovering his mouse over different buildings to show who’d crafted them. “And who knows where these people are and where they live? But that’s the kind of cool thing about it. People are able to communicate through this tool, where they can share all this information.” Google Earth’s relative openness—and its value as a creative showcase for one’s 3-D-modeling skills—turned out to be a tempting invitation to contribute, even though Google is clearly a for-profit entity. (In 2013, Google launched an even faster way of generating buildings—by using satellite photo data—and retired the building-maker tool, though it kept many of the buildings created by contributors.)
Still, because openness is most natural in amateur work, I suspect the leading edge of collective thinking—as with Wikipedia or Linux—will always emerge in the amateur world. If you want to see the future of collective thinking, don’t watch what Fortune 500 firms are doing. Watch what fan fiction writers are doing or what activists are doing. Or even watch how smart individuals do it—the ones who cultivate broad, diverse networks of friends or followers online.
The potential of collective thinking is only going to grow. Each new tool for expression, each new vehicle for talking to one another, opens up new potential forums for collaboration. Some regard this as alarming: Critics have complained that the online “hive mind” is dehumanizing. And it’s true that thinking of people as bees in a hive, devoid of agency, is rather depressing. But this is precisely why the hive metaphor isn’t all that accurate. Humans are not ants, as the philosopher Pierre Lévy noted in his 1994 book Collective Intelligence. We participate in larger groupings when there’s something there that enhances our individual humanity. The collaborative thought projects that succeed are the ones where each act of participation, no matter how small, excites and rehumanizes us. This is why Wikipedia succeeds. Those who make an edit or participate in a talk-page discussion transform their sense of self, becoming creators of knowledge.
Personally, I think there’s a better metaphor for collaborative thinking: Sherlock Holmes. Arthur Conan Doyle’s detective is famously brilliant. But he’s also famously bored. When he doesn’t have a complex case, something so fiendishly difficult that it engages all of his electric intellect, the boredom drives him nuts. He shoots cocaine (“a seven-per-cent solution”) and begs for the universe to send him a case complicated enough to engage his analytic abilities.
“My mind rebels at stagnation,” he tells Watson in The Sign of the Four. “Give me problems, give me work, give me the most abstruse cryptogram, or the most intricate analysis, and I am in my own proper atmosphere. I can dispense then with artificial stimulants. But I abhor the dull routine of existence. I crave for mental exaltation.” Or in The Adventure of Wisteria Lodge: “My mind is like a racing engine, tearing itself to pieces because it is not connected up with the work for which it was built.”
This is what the enormous, latent collaborative intelligence of humanity is like: one big Sherlock Holmes, craving problems that suit its peculiar powers. Like the networked video-game players, it’s hungry for harder puzzles. We have to learn how to design them.