by Clay Shirky
AT&T was right to be skeptical about community; it has not historically been a good guarantor of longevity. The fact that shared interest can now create that longevity is what makes the current change historic. This is the secret of the open source ecosystem and, by extension, of all the large-scale and long-lived forms of sharing, collaborative work, and collective action now being tried. Because anyone can try anything, the projects that fail, fail quickly, but the people working on those projects can migrate just as quickly to the things that are visibly working. Unlike the business landscape, where companies have an incentive to hide both successes (for reasons of competitive advantage) and failures (to forestall any perception of weakness), open source projects advertise their successes and get failure for free. This arrangement allows the successes to become host to a community of sustained interest.
What the open source movement teaches us is that the communal can be at least as durable as the commercial. For any given piece of software, the question “Do the people who like it take care of each other?” turns out to be a better predictor of success than “What’s the business model?” As the rest of the world gets access to the tools once reserved for the techies, that pattern is appearing everywhere, and it is changing society as it does.
CHAPTER 11
PROMISE, TOOL, BARGAIN
There is no recipe for the successful use of social tools. Instead, every working system is a mix of social and technological factors.
Every story in this book relies on a successful fusion of a plausible promise, an effective tool, and an acceptable bargain with the users. The promise is the basic “why” for anyone to join or contribute to a group. The tool helps with the “how”—how will the difficulties of coordination be overcome, or at least be held to manageable levels? And the bargain sets the rules of the road: if you are interested in the promise and adopt the tools, what can you expect, and what will be expected of you? The interaction of promise, tool, and bargain can’t be used as a recipe, because the interactions among the different components is too complex. Like the weather, the complex interaction of the various forces makes the results only partially predictable.
The order of promise, tool, and bargain is also the order in which they matter to the success of any given group. Making a promise that enough people believe in is the basic requirement; the promise creates the basic desire to participate. Then come the tools. After getting the promise right (or right enough), the next hurdle is figuring out which tools will best help people approach the promise together. Wikis make arriving at shared judgment easier than hosting a discussion, while e-mail has the opposite set of characteristics, so getting the tools right matters to the kind of interactions the group will rely on. Then comes the bargain. Tools don’t completely determine behavior; different mailing lists have different cultures, for example, and these cultures are a result of an often implicit bargain among the users. One possible bargain for a mailing list is: “We expect politeness of one another, and we rebuke the impolite.” Another, very different bargain is: “Anything goes.” You can see how these bargains would lead to very different cultures, even among groups using the same tools, yet both patterns exist in abundance. A successful bargain among users must be a good fit for both the promise and the tools used. Taken together, these three characteristics are useful for understanding both successes and failures of groups relying on social tools.
The promise is the essential piece, the thing that convinces a potential user to become an actual user. Everyone already has enough to do, every day, and no matter what you think of those choices (“I would never watch that much TV,” “Why are they at work at ten p.m.?”), those choices are theirs to make. Any new claim on someone’s time must obviously offer some value, but more important, it must offer some value higher than something else she already does, or she won’t free up the time. The promise has to hit a sweet spot among several extremes. The original promise of Voice of the Faithful was neither too mundane (“Let’s blow off some steam about abusive priests”) nor too disrespectful (“Let’s demolish the Church”). Instead, its message balanced loyalty with anger—“Keep the faith, change the church.” Just right, at least for purposes of recruiting. Similarly, the original message inviting people to work on the Linux operating system was neither too provisional (“Let’s try to see if we can come up with something together”) nor too sweeping (“Let’s create a world-changing operating system”). Instead, Linus’s proposal was modest but interesting—a new but small operating system, undertaken principally as a way to learn together. Just right.
The implicit promise of any given group matters more than any explicit one, which is to say that the stated rationale of the group is not necessarily the lived one. The explicit promise of pro-anorexia sites is to be able to get and remain un-healthily thin, but when you read the material posted on those sites, you can see that the actual promise is something more like: “Someone will pay attention to you.” Much of the material on these sites is written from the perspectives of girls who have recovered from anorexia: as in other clubs, the pleasure of other people’s company is often as important as, and sometimes more important than, the excuse for getting together in the first place.
The problem of getting the promise right is unlike traditional marketing, because most marketing involves selling something that will be made for the listeners rather than by them. “Buy Cheesy Poofs” is a different message from “Join us, and we will invent Cheesy Poofs together.” This second kind of message is more complicated, because of something called the paradox of groups. The paradox is simple—there can be no group without members (obviously), but there can also be no members without a group, because what would they be members of? Single-user tools, from word-processing software to Tetris, have a simple message for the potential user: if you use this, you will find it satisfying or effective or both. With social tools, the group is the user, so you need to convince individuals not just that they will find the group satisfying and effective but that others will find it so as well; no matter how appealing the promise, there’s no point in being the only user of a social tool. As a result, users of social tools are making two related judgments: Will I like using this tool or participating in this group? Will enough other people feel as I do to make it take off?
The larger the number of users required, the harder the group is to get going, because the potential users will (rightly) be more skeptical that enough users will join to make it worth their while. (An empty restaurant has the same catch-22 in attracting diners.) There are several strategies for handling this problem. The most obvious one is to make joining easy, in order to make the promise seem within reach. Kate Hanni’s Flyers Rights group made the basic action (signing the petition) quite simple and reserved more complicated actions (like calling Congress or talking to the media) for more committed members. Other strategies include creating personal value for the individual users, allowing the social value to manifest only later. Joshua Schachter’s service for bookmarking and tagging webpages, called del.icio.us, serves as a personal archive of webpages; the value that accrues from aggregating the group’s view of the Web is optional for any given user, but enough people have taken advantage of that value to cause the service to grow dramatically.
Another common strategy is to subdivide the community, in a Small World pattern, so that small but densely connected clusters of people have value even before the service grows large. LiveJournal, the weblogging platform, got a lot of its early growth from clusters of high school students joining at the same time. Though LiveJournal offered more value as it grew bigger (more people to meet, more possible groups to join), it offered enough value to small groups to be able to grow large. (MySpace did something similar during its early growth.) And sometimes good old-fashioned hosting helps bridge the gap, making a promise seem plausible even while users are too few. Part of the promise of Flickr, the photo-sharing platform, was that the public could see your photos. (Flickr made the sharing of photos
the default option, though users could turn it off.) Yet the attraction of such photos required an audience, and the logical place to get that audience was from among other Flickr users. Like the proverbial stone soup, the promise would be achieved only if everyone participated, and like the soldiers who convince the townspeople to make the stone soup, the only way to hold the site together before it reached critical mass was through personal charisma. Caterina Fake, one of the founders of Flickr, said she’d learned from the early days that “you have to greet the first ten thousand users personally.” When the site was small, she and the other staffers would not just post their own photos but also comment on other users’ photos, like a host circulating at a party. This let the early users feel what it would be like to have an appreciative public, even before such a public existed.
Of course, Fake couldn’t realistically promise Flickr users that their photos would be admired—most photos are in fact quite dull, on Flickr as elsewhere. What she could tell them was that if they worked to produce admirable photos, they had a chance at finding an audience. The promise of Wikipedia is similarly that you have a chance at having your contributions to an article last, and the promise of weblogs is that you have a chance of finding people who want to read your writing. In the end, because the value of these groups is derived from the participation of the group, the promise is more of a challenge than a guarantee.
Tools
After the complexity of the question of figuring out the promise of a given group, the question of determining which tools to use seems as if it should be easy. Here again context complicates things. There is no such thing as a generically good tool; there are only tools good for particular jobs. Contrary to the hopes of countless managers, technology is not an infinitely elastic piece of fabric that can be stretched to cover any situation. Instead, a good social tool is like a good woodworking tool—it must be designed to fit the job being done, and it must help people do something they actually want to do. If you designed a better shovel, people would not rush out to dig more ditches.
One surprising ramification of this “goodness of fit” argument is that when you improve the available tools, you expand the number of plausible promises in the world. Linus Torvalds’s original promise for Linux seems small in retrospect, but stated baldly—“Let’s get a bunch of people all over the world to write incredibly complex software without anyone getting paid”—the proposal would have seemed utterly mad. (Many people treated Linux that way for years, in fact.) Richard Stallman’s more managed methods of creating software seemed better than Torvalds’s, because up to that point they had been better. In the early 1990s Torvalds’s proposal hit the forward edge of what social tools made plausible, and as the tools got better, the size of what was plausible grew. The social tools that the Linux community adopted were like a trellis for vines—they didn’t make the growth possible, but they supported and extended that growth in ways that let them defy gravity.
We are living in the middle of a huge increase in the number of available tools: the launch of Twitter, the text-messaging tool, happened during the writing of this book. Given this profusion, can we say anything useful about the future social landscape? Yes, but only by switching focus from the individual tools themselves to the kinds of groups the tools are expected to support. Two of the most critical questions are “Does the group need to be small or large?” and “Does it need to be short-lived or long-lived?” Two either/or questions mean four possible combinations; a flash mob is a small short-lived group, while the people contributing to Linux comprise a large and long-lived one, and so on.
The core characteristic of small groups is that their members can interact more tightly with one another, because social density is easier to support in small groups than in large ones (the result of Birthday Paradox math, and part of what drives the Small World pattern). Small groups are thus better conversational environments than large ones and find it easier to engage in convergent thinking, where everyone comes to agree on a single point of view. This is one of the things social tools don’t change about group life—small groups are more effective at creating and sustaining both agreement and shared awareness.
The core characteristics of large groups are the inverse. People have to be less tightly connected, on average, to one another. As a result, such groups are better able to produce what James Surowiecki has called “the wisdom of crowds.” In his book of that name he identified the ways distributed groups whose members aren’t connected can often generate better answers, by pooling their knowledge or intuition without having to come to an agreement. We have many ways of achieving this kind of aggregation, from market pricing mechanisms to voting to the prediction markets Surowiecki champions, but these methods all have two common characteristics: they work better in large groups, and they don’t require direct communication as the norm among members. (Indeed, in the case of markets, such communication is often forbidden, on the grounds that small clusters of collaborators can actually pervert the workings of the large system.)
Small and large are relative rather than absolute. In a home, dinner for twelve is (usually) a large group, while in a school a class of a dozen students is small. The coordination issues of dinner are more intense than the coordination issues of discussion. Similarly, a hundred people turning up at a Meetup is a large group, while a hundred people turning out for a political rally is a small one. Whatever the issues of relative size, however, the absolute issue remains—larger groups have looser ties.
Making a promise without having a way to deliver on it isn’t plausible, by definition. Tools are tied to the modes of group interaction they need to support. You can see how this is so by imagining switching tools among different groups. The Flyers Rights group and the Egyptian prodemocracy activists both wanted to change the laws of their respective countries. The Flyers Rights group worked deliberately, gathering support over weeks using weblogs and online petition forms. The activists in Cairo used blogs, but some of this activism took place at faster speeds, coordinating in the streets of Cairo using Twitter. Imagine trying to force the Flyers Rights people to use Twitter, while limiting the dozen or so activists trying to save their friend Malek to weblogs. Both groups would have failed. Twitter would have annoyed the people who were happy to sign a petition—the online petition form is a slow-motion but high-visibility tool. Likewise, Alaa Abd El Fattah and his friends could never have coordinated in the streets of Cairo using weblogs—they needed a fast but low-visibility tool like Twitter.
By understanding these two basic constraints of group action—number of people involved and duration of interaction—any given tool, new or familiar, can be analyzed for goodness of fit. And of course a single service can offer more than one tool and thus support more than one form of interaction. The Windows programmers who hang out on the #winprog chat channel use a tool that supports conversational interaction, but they put their collected group wisdom on a set of webpages, including a FAQ, a list of frequently asked questions, with answers. A FAQ is a social document, representing accumulated wisdom about the commonest questions that arise within a group. The rate at which the FAQ is updated is much slower than the rate at which conversation happens on the chat channel, allowing the community to operate at more than one speed. Similarly, the group that congregates to talk about a particular Wikipedia article may be quite small, while the contributor base of Wikipedia as a whole is enormous, allowing Wikipedia to operate at more than one scale.
Perhaps most important, new tools are not always better. New tools, in fact, start with a huge social disadvantage, which is that most people don’t use them, and whenever you have a limited pool from which potential members can be drawn, you limit the social effects. In addition, every social tool is surrounded by an ocean of practice, which helps dictate its use. When the denizens of Bronze, an online discussion group for the TV show Buffy the Vampire Slayer, learned that the TV network was no longer going to support their community, they rallied together, raising eno
ugh money to commission new software in a new location, like a hermit crab requisitioning a new shell. When they hired a firm to create the new tool, they had one simple request—no major changes. The old tool, which they had gotten used to, was absolutely bare-bones, and they sensed that if the new tool added complicated features, the community would suffer. Instead, they requested (and got) something that looks laughably simple by the standards of more recent software. Their intuition turned out to be correct: the community survived the move from the old location to the new one, which they dubbed Bronze:Beta.
Many of the stories in this book involve the most mundane tools: e-mail lists and discussion groups have been around since the 1970s, and even many of the newer tools, like weblogs and wikis, are already a decade old. The most profound effects of social tools lag their invention by years, because it isn’t until they have a critical mass of adopters, adopters who take these tools for granted, that their real effects begin to appear.
Bargains
The bargain comes last, because it matters only if there is a promise and a set of tools that are already working together. The bargain is also the most complex aspect of a functioning group, in part because it is the least explicit aspect and in part because it is the one the users have the biggest hand in creating, which means it can’t be completely determined in advance. The need for a bargain gets back to the most basic issues of group effort—transaction costs. A bargain helps clarify what you can expect of others and what they can expect of you. Imagine you are traveling to a foreign country and are planning to drive while there. Do you want to drive on the right side of the road or the left? The answer, of course, is that right or left is the wrong question—you want to drive on whatever side of the road everyone else does; synchronization with the natives is itself a value. (When in Rome . . .) The same is true for mediated groups; there are many different ways social expectations can be worked out, but as with the rules of the road, what matters is that there is a way and that everyone knows what it is.
