Here Comes Everybody

by Clay Shirky

  One very public illustration of the difference between bridging and bonding capital came in the form of Howard Dean’s presidential campaign. At the end of 2003 Dean had the best-funded, best-publicized bid to be the Democratic nominee. Dean was so widely understood to be in the lead that the inevitability of his victory was a broad topic of discussion. Even the people who disputed this inevitability burnished the idea; no one bothered disputing the inevitability of any of the other candidates. And yet Dean’s campaign was unsuccessful. It did many of the things successful campaigns do—it got press coverage and raised money and excited people and even got potential voters to aver to campaign workers and pollsters that they would vote for him when the time came. When the time came, however, they didn’t. The campaign never succeeded in making Howard Dean the first choice of any group of voters he faced.

  The Dean campaign brilliantly conveyed a message to its supporters, particularly its young ones, that their energy and enthusiasm could change the world. Some of this message was conveyed by design, but much of it was a function of people looking for something, finding it in Dean, and then using tools like Meetup and weblogs to organize themselves. The Dean campaign was unequaled in creating bonding capital among its most ardent supporters. They gained a sense of value just from participating; and in the end the participation came to matter more than the goal (a pretty serious weakness for a vote-getting operation). The pleasure in working on the Dean campaign was in knowing that you were on the right side of history; the campaign’s brilliant use of social tools to gather the like-minded further fed that feeling. It is natural for a campaign attracting so many eager young people to oversell them on the effect they’ll have, when the truth is so rough: you’ll work eighty-hour weeks while sleeping on someone’s sofa, and in the end your heroic contribution will be a drop in the bucket of what’s needed. So a little pep talk now and again can’t hurt.

  But a campaign can go too far. In this respect, too far is when people believe that believing is enough, without factoring in the difference between the passionate few who run the campaign and the barely interested many who actually vote. Voting, the heart of the matter, is both dull and depressing. Standing around an elementary school cafeteria is not a great way to feel like your energy and excitement are going to change the world, because the math of the voting booth undermines any sense of inevitability: everyone in line not voting for Dean cancels your vote. The Dean campaign had accidentally created a movement for a passionate few rather than a vote-getting operation.

  Bonding capital tends to be more exclusive and bridging capital more inclusive. In Small World networks bonding tends to happen within the clusters, while bridging happens between clusters. The Dean campaign was great at doing everything a campaign can do with bonding capital—gathering ardent supporters and raising millions in funds—but getting people to vote for the candidate required bridging capital, reaching out to people outside the charmed inner circle. The Pro-Ana groups also have and use bonding capital well—members are relatively homogenous with regard to age and class and completely so with regard to gender. Meetup relies on (and produces) bonding and bridging capital along a spectrum, depending on the group. A Meetup group for Ping-Pong would produce bridging capital (anyone can join the group, regardless of age, class, or gender), while Black Stay at Home Moms relies on homogeneity as part of the appeal.

  The effects of homophily touch every social system; technology doesn’t free us from social preferences or prejudices. As danah boyd, the great scholar of social networks (who doesn’t capitalize her name), has noted, the populations of MySpace and Facebook, two large social networking sites, mirror divisions in American class structure. Facebook started as a site for college students, so when it opened its virtual doors to high school students as well, it was framed as being for the college-bound, while MySpace remained, in boyd’s words, a home for “the kids who are socially ostracized at school because they are geeks, freaks, or queers.” Even our modest preferences for bonding can lead to large-scale divisions of this sort.

  Perhaps the most significant effect of our new tools, though, lies in the increased leverage they give the most connected people. The tightness of a large social network comes less from increasing the number of connections that the average member of the network can support than from increasing the number of connections that the most connected people can support.

  Bridging Capital, 24/7

  Joi Ito is variously an investor, a writer, a hardcore gamer, and a member of the board of scores of companies and nonprofits. His address book contains several thousand names. He is on the road constantly; in 2005 he traveled so much that his average speed was fifty miles per hour for that year. Ito is also an inveterate adopter of new technologies; he tries a remarkable number of social and organizational tools every year and sticks with the ones that make sense to him. One of the tools he adopted a few years ago was internet relay chat, or IRC, an old tool (created in 1988) that creates a real-time chat room called a channel. Everyone using a particular IRC channel can talk to everyone else on that channel. (I use “talk to” here in its colloquial sense of “type quickly to.”) Channels on IRC are like instant message or text message conversations, but instead of being people-centric, they are topic-centric. A channel’s name gives the users some sense of what the channel’s denizens are talking about or at least share in common. Some channels are long-lived—there is a decades-old channel called #hottub (IRC channels always begin with a # sign), mainly dedicated to flirty talk among bored college students the world over. Other IRC channels are transient—a dozen separate IRC channels were set up during the low-speed police chase of O. J. Simpson in 1994, the participants speculating on the result of the chase as it was happening.

  In 2004, Ito set up an IRC channel called #joiito, where his friends and contacts could congregate and talk. It was meant to be, as he puts it, “not my place, but a semipublic place, where I could be a host.” He used his name both because he is recognized in so many communities (this isn’t vanity, just observation—a Web search for “Joi Ito” brings up nearly a million results) and because he wanted to be able to exert some sort of moral suasion over the proceedings. If the IRC channel bore his name, he had a better chance of enforcing civil behavior. The channel quickly grew to have a hundred or so people logged in at the same time. Most of these people were not talking most of the time—they would leave themselves logged in even if they were not paying attention to the channel, or even present at their computers. But their presence ensured relatively regular conversation among the channel’s denizens.

  One of the regulars, a programmer named Victor Ruiz, wrote a piece of software called jibot (short for “#joiito bot,” where a “bot” is an interactive program). Jibot would monitor the channel and answer specially formatted questions, including looking up words in a custom dictionary. One of the regulars brought in a new user, Jeannie Cool, who became an unofficial hostess. She adopted jibot as a social tool by entering in “definitions” tied to the other users’ names. Seeing this, Kevin Marks, another regular, modified jibot to make an announcement of this “definition” whenever anyone logged in (a function called heralding). For instance, when the user mmealling shows up, jibot would post the phrase “mmealling is Michael Mealling. He lives in Atlanta, GA.” This design, simple as it was, helped move #joiito from a place mainly set up for bonding capital (geeks who knew Joi) into a place that produced bridging capital (people who knew geeks who knew Joi). The jibot, in other words, made #joiito more like a bowling league, which you could join without needing to know most of the members.

  The intersection between social networks and electronic networks here is simple in its elements but complex in its results, partly because so many feedback loops are involved. Joi adopted IRC because it was a good way to offer a standing site for interaction among people he knew. The people who gathered there got to know one another better through their interactions, and as with any successful community, new members were
invited in. These new members didn’t have all the context the original members did, and to meet this need, Ruiz used his technical skills to customize software to create heralding. Critically, the use of jibot as a social tool preceded the heralding function—the reworking of the software was a reflection of behavior, not the other way around. To this day #joiito has about eighty people logged in at any given time. The existence of the channel allows Joi to create a persistent environment where people who know him can meet one another, even if he is not present. Once the channel achieved a kind of social stability, he logged in less and less often; there are people on #joiito around the clock and around the world who do not need a lot of Joi Ito to make it happen. (Dodgeball similarly relies on its users’ social capital to help broker introductions without requiring that they be present.) In this way #joiito is an instantiation of Joi’s role as a connector; if everyone were to start their own IRC channel, no one would ever talk to anyone else, as they’d all be in their own solo spaces. A channel called #joiito makes social sense—we rely on Joi to provide irreplaceable bridging capital—while one called #jrandomuser wouldn’t make sense. Like a bar or a café, #joiito hosts the kinds of informal conversations of which social capital is made; unlike a bar or a café, it costs nothing to run.

  Joi’s IRC channel is unusual, but the ability of one person or a small group to create this kind of social value is not. Another IRC channel, #winprog, is a hangout for some of the world’s most talented Windows programmers. Like many geek hangouts, the channel is a brutal technical meritocracy (rule number one of the channel: “No whining”), but for people serious about Windows programming, #winprog is invaluable, both as a source of information and as a way for serious programmers to get integrated into a community of practice. As a source of both information and camaraderie, #winprog is satisfying and effective for its members.

  Similarly, Howard Forums is a Web discussion board founded by Howard Chui, a computer programmer who became obsessive about mobile phones. Chui founded Howard Forums after fielding technical questions from a number of readers to his mobile phone weblog; he reasoned that putting his readers in touch with one another would be easier than trying to answer all their questions himself. The intuition proved correct; less than five years after its founding, the site gets half a billion page-views a year on incredibly detailed subjects, such as customizing specific brands of phones or the merits of various mobile networks. The information produced is so good that engineers at mobile phone companies will sometimes refer customers to it when they have a particularly complex question. Despite the fact that Howard Forums is not an official part of any mobile company, the quality of the technical information there is outstanding, a product of the community’s passion for (or obsession with) phones.

  Tim O’Reilly, the publisher and conference organizer, founded the conference FOO Camp (Friends of O’Reilly). This conference starts from the invite list—gather a hundred interesting people—and lets them work out the schedule and content of the conference (on a wiki, of course). All these forms suggest that structured aggregation of individual interests and talents can create a kind of value that is hard to replicate with ordinary institutional forms, and impossible to replicate at such low cost.

  It’s Not How Many People You Know, It’s How Many Kinds

  In one of the more evocatively titled papers in the history of social science, “The Social Origins of Good Ideas,” Ronald Burt of the University of Chicago detailed his research into the relationship among social capital, social structure, and good ideas. The research method was simple and relatively direct (though interpreting the data wasn’t). Burt looked at a major U.S. electronics firm that was undergoing a change in management in 2001, and he got the new management to agree to participate in an experiment. They would ask the managers responsible for the company’s supply chain to submit ideas for improving the business and to reveal who else in the firm they’d discussed the idea with, if anyone. This experiment provided a good context for observing the social relations among the company’s staff, since the employees responsible for the supply chain were often isolated from the rest of the company. Two of the new managers would then rank the ideas on a scale of 1 to 5; they also had the option to reject an idea outright, if it was “too local in nature, incomprehensible, vague, or too whiny,” in the words of one of the senior managers. (Note the resistance to whining, as with #winprog—one of the common requirements to group participation is setting aside purely personal issues in the group setting.)

  The essence of Burt’s thesis comes down to a linked pair of observations. First, most good ideas came from people who were bridging “structural holes,” which is to say people whose immediate social network included employees outside their department. Second, bridging these structural holes was valuable even when other variables, such as rank and age (both of which correlate for higher degrees of social connection), were controlled for. Note that this experiment was a test for bridging capital, not mere sociability—the highest percentage of good ideas came from people whose contacts were outside their own department. On the other hand, managers who were highly connected, but only to others in their department, had ideas that were not ranked as highly. Bridging predicted good ideas; lack of bridging predicted bad ones.

  In Burt’s analysis, a dense social network of people in the same department (and who were therefore likely to be personally connected to one another) seemed to create an echo-chamber effect. The new managers rejected ideas drawn from this pool with disproportionate frequency, often on the grounds that the ideas were too involved in the minutiae of that particular department and provided no strategic advantage for the company as a whole.

  Nor was this experiment a test for intellect. As Burt puts it in the paper:

  People whose networks span structural holes have early access to diverse, often contradictory, information and interpretations which gives them a good competitive advantage in delivering good ideas. People connected to groups beyond their own can expect to find themselves delivering valuable ideas, seeming to be gifted with creativity. This is not creativity born of deep intellectual ability. It is creativity as an import-export business. An idea mundane in one group can be a valuable insight in another.

  Burt found that bridging capital puts people at greater risk of having good ideas (his phrase) than do any individual traits. For something like supply chain management, it’s easy to see why this might be so—the department handling that function was separated from the rest of the electronics business and was not seen as a core function. The converse was also true; if the proposer of an idea was talking only within his or her own department, the idea was much likelier to be parochial. It seems so simple—just mix people up and sit back and watch the good ideas roll in. There must be a catch.

  And there is. Even when the judicious use of social connections increases the proportion of good ideas, most ideas are still bad. It’s not enough to find some way to increase the successful ideas. Some way needs to be found to tolerate the failures too.

  CHAPTER 10

  FAILURE FOR FREE

  The logic of publish-then-filter means that new social systems have to tolerate enormous amounts of failure. The only way to uncover and promote the rare successes is to rely, yet again, on social structure supported by social tools.

  The Stay at Home Moms (see Chapter 8) have pressed the generic capabilities of Meetup into service to create a sense of local community that is otherwise hard to arrange in a physically dispersed culture. It’s obvious why they would find Meetup valuable. Less obvious but as least as remarkable is how that particular group came to be in the first place. Every Meetup group navigates the tension between specificity and size. A Meetup group perfectly fit to an individual (bald fathers of two in Brooklyn who teach at NYU and like bagpipe music) would have exactly one member, while a Meetup that included huge numbers of potential members (parents, or TV watchers, or residents of Atlanta) would provide little in the way of commonality or con
versational fodder: “So, you watch TV too, huh?” The ideal group exists in some equipoise between specific and generic. The Stay at Home Moms group fits that description well enough that it is more popular than all the other parenting groups and one of the most popular Meetup categories overall.

  Even accepting that Stay At Home Moms groups exist at some optimum point between size and specificity, there’s still a mystery about its formation: How did Meetup know that the group would be as appealing as it was? Most of the people who work at Meetup are overeducated, undermarried urbanites who face a completely different set of problems than do the North Charlotte Stay at Home Moms. How could they have known that SAHM groups would be such a hit?

  They didn’t. To have predicted such a thing, the employees of Meetup would have needed research about the changing face of American communities, current trends in self-definition of mothers, interactions among suburbanites, and so on; demographics, psychology, sociology. Even if someone had told them that Stay at Home Moms was a good idea for a Meetup group, the staff might have been loath to propose such a thing. Coming from a bunch of single urbanites, it might have seemed patronizing, to say nothing of polarizing. The staff might have become a target for political protest by people upset about the exclusivity implied by the name. Meetup could not have gathered enough information to understand which parenting groups to suggest in the first place, could not have picked a winner even if they’d had all that information, and could not have launched the winner even if they’d been able to pick one, because of the potential negative reaction.