Like many debates from the annals of urban studies, the Mumford/Jacobs exchange over the “climax stage” of city life mirrors recent developments in the digital realm, as Web-based communities struggle to manage the problems of runaway growth. The first generation of online hangouts—dial-up electronic bulletin boards like ECHO or the Well—were the equivalent of those Italian hill-towns: lively, innovative, contentious places, but also places that remained within a certain practical size. In their heyday before the Web’s takeoff, both services hovered around five thousand members, and within that population, community leaders and other public characters naturally emerged: the jokers and the enablers, the fact checkers and the polemicists. These characters—many of them concealed behind playful pseudonyms—served as the equivalent of Jacobs’s shopkeepers and bartenders, the regular “eyes on the street” that give the neighborhood its grounding and its familiarity.
These online communities also divided themselves into smaller units organized around specific topics. Like the trade-specific clusters of Savile Row and the Por Santa Maria, these divisions made the overall space more intelligible, and their peculiarities endowed each community with a distinctive flavor. (For the first few years of its existence, the Grateful Dead discussion area on the Well was larger than all the other areas combined.) Because each topic area attracted a smaller subset of the overall population, visiting each one felt like returning to an old block in a familiar part of town, and running into the same cast of characters that you had found there the last time you visited.
ECHO and the Well had a certain homeostatic balance in those early years—powerfully captured in Howard Rheingold’s book The Virtual Community—and part of that balance came from the community’s own powers of self-organization. But neither was a pure example of bottom-up behavior: the topic areas, for instance, were central-planning affairs, created by fiat and not by footprints; both communities benefited from the strong top-down leadership of their founders. That their overall populations never approached a “climax stage” reflected the slow modem-adoption curve of the general public, and the limited marketing budgets at both operations. More important, the elements of each community that did self-regulate had little to do with the underlying software. Anyone who spent any time on those services in the early nineties will tell you that community leaders and other recognizable figures emerged, but that status existed only in the perceptions of the users themselves. The software itself was agnostic when it came to status, but because the software brought hominid minds together—minds that are naturally inclined to establish hierarchies in social relationships—leaders and pariahs began to appear. The software did recognize official moderators for each discussion area, but those too were appointments handed down from above; you applied to the village chieftain for the role that you desired, and if you’d been a productive member of the society, your wish might be granted. Their were plenty of unofficial leaders, to be sure—but where the code was concerned, the only official moderators came straight from the top.
This mix of hierarchy and heterarchy was well suited to ECHO’s and the Well’s stage of growth. At five thousand members, the community was still small enough to be managed partially from above, and small enough that groups and recognizable characters naturally emerged. At that scale, you didn’t need to solve the problem of self-regulation with software tools: all you needed was software that connected people’s thoughts—via the asynchronous posts of a threaded discussion board—and the community could find its own balance. If something went wrong, you could always look to the official leaders for guidance. But even in those heady early days of the virtual community, the collective systems of ECHO and the Well fell short of achieving real homeostasis, for reasons that would become endemic to the next generation of communities then forming on the Web itself.
A threaded discussion board turns out to be an ideal ecosystem for that peculiar species known as the crank—the ideologue obsessed with a certain issue or interpretive model, who has no qualms about interjecting his or her worldview into any discussion, and apparently no day job or family life to keep him from posting voluminous commentary at the slightest provocation. We all know people like this, the ones grinding their ax from the back of the seminar room or the coffee shop: the conspiracy theorist, the rabid libertarian, the evangelist—the ones who insist on bringing all conversations back to their particular issue, objecting to any conversation that doesn’t play by their rules. In real life, we’ve developed a series of social conventions that keep the crank from dominating our conversations. For the most pathological cases, they simply don’t get invited out to dinner very often. But for the borderline case, a subtle but powerful mechanism is at work in any face-to-face group conversation: if an individual is holding a conversation hostage with an irrelevant obsession, groups can naturally establish a consensus—using words, body language, facial expressions, even a show of hands—making it clear that the majority of the group feels their time is being wasted. The face-to-face world is populated by countless impromptu polls that take the group’s collective pulse. Most of them happen so quickly that we don’t even know that we’re participating in them, and that transparency is one reason why they’re as powerful as they are. In the face-to-face world, we are all social thermostats: reading the group temperature and adjusting our behavior accordingly.
Some of those self-regulatory social skills translate into cyberspace—particularly in a threaded discussion forum or an e-mail exchange, where participants have the time and space to express their ideas in long form, rather than in the spontaneous eruptions of realtime chat. But there is a crucial difference in an environment like ECHO or the Well—or in the discussion areas we built at FEED. In a public discussion thread, not all the participants are visible. A given conversation may have five or six active contributors and several dozen “lurkers” who read through the posts but don’t chime in with their own words. This creates a fundamental imbalance in the system of threaded discussion and gives the crank an opportunity to dominate the space in a way that would be much more difficult off-line. In a threaded discussion, you’re speaking both to the other active participants and to the lurkers, and however much you might offend or bore your direct interlocutors, you can always appeal to that silent majority out there—an audience that is both present and absent at the same time. The crank can cling to the possibility that everyone else tuning in is enthralled by his prose, while the active participants can’t turn to the room and say, “Show of hands: Is this guy a lunatic or what?”
The crank exploits a crucial disparity in the flow of information: while we conventionally think of threaded discussions as two-way systems, for the lurkers that flow follows a one-way path. They hear us talking, but we hear nothing of them: no laughs, no hisses, no restless stirring, no snores, no rolling eyeballs. When you factor in the lurkers, a threaded discussion turns out to be less interactive than a traditional face-to-face lecture, and significantly less so than a conversation around a dinner table, where even the most reticent participants contribute with gestures and facial expressions. Group conversations in the real world have an uncanny aptitude for reaching a certain kind of homeostasis: the conversation moves toward a zone that pleases as much of the group as possible and drowns out voices that offend. A group conversation is a kind of circuit board, with primary inputs coming from the official speakers, and secondary inputs coming from the responses of the audience and other speakers. The primary inputs adjust their signal based on the secondary inputs of group feedback. Human beings—for reasons that we will explore in the final section—are exceptionally talented at assessing the mental states of other people, both through the direct exchanges of spoken language and the more oblique feedback mechanisms of gesture and intonation. That two-way exchange gives our face-to-face group conversations precisely the flexibility and responsiveness that Wiener found lacking in mass communications.
I suspect Wiener would immediately have understood the virtual communit
y’s problem with cranks and lurkers. Where the Flowers affair was a case of runaway positive feedback, the tyranny of the crank results from a scarcity of feedback: a system where the information flows are unidirectional, where the audience is present and at the same time invisible. These liabilities run parallel to the problems of one-way linking that we saw in the previous chapter. Hypertext links and virtual communities were supposed to be the advance guard of the interactive revolution, but in a real sense they only got halfway to the promised land. (Needless to say, the ants were there millions of years ago.) And if the cranks and obsessive-compulsives flourish in a small-scale online community of several thousand members, imagine the anarchy and noise generated by a million community members. Surely there is a “climax stage” on that scale where the online growth turns cancerous, where the knowable community becomes a nightmare of overdevelopment. If feedback couldn’t help regulate the digital villages of early online communication, what hope can it possibly have on the vast grid of the World Wide Web?
*
The sleepy college town of Holland, Michigan, might seem like the last place you’d expect to generate a solution for the problem of digital sprawl, but the Web has never played by the rules of traditional geography. Until recent years, Holland had been best known for its annual tulip festival. But it is increasingly recognized as the birthplace of Slashdot.org—the closest thing to a genuinely self-organizing community that the Web has yet produced.
Begun as a modest bulletin board by a lifetime Hollander named Rob Malda, Slashdot came into the world as the ultimate in knowable communities: just Malda and his friends, discussing programming news, Star Wars rumors, video games, and other geek-chic marginalia. “In the beginning, Slashdot was small,” Malda writes. “We got dozens of posts each day, and it was good. The signal was high, the noise was low.” Before long, though, Slashdot floated across the rising tsunami of Linux and the Open Source movement and found itself awash in thousands of daily visitors. In its early days, Slashdot had felt like the hill towns of ECHO and the Well, with strong leadership coming from Malda himself, who went by the handle Commander Taco. But the volume of posts became too much for any single person to filter out the useless information. “Trolling and spamming became more common,” Malda says now, “and there wasn’t enough time for me to personally keep them in check and still handle my other responsibilities.”
Malda’s first inclination was to create a Slashdot elite: twenty-five handpicked spam warriors who would sift through the material generated by the community, eliminating irrelevant or obnoxious posts. While the idea of an elite belonged to a more hierarchical tradition, Malda endowed his lieutenants with a crucial resource: they could rate other contributions, on a scale of -1 to 5. You could browse through Slashdot.org with a “quality filter” on, effectively telling the software, “Show me only items that have a rating higher than 3.” This gave his lieutenants a positive function as well as a negative one. They could emphasize the good stuff and reward users who were productive members of the community.
Soon, though, Slashdot grew too large for even the elites to manage, and Malda went back to the drawing board. It was the kind of thing that could only have happened on the Web. A twenty-two-year-old college senior, living with a couple of buddies in a low-rent house—affectionately dubbed Geek House One—in a nondescript Michigan town, creates an intimate online space for his friends to discuss their shared obsessions, and within a year fifty thousand people each day are angling for a piece of the action. Without anything resembling a genuine business infrastructure, much less a real office, Malda needed far more than his twenty-five lieutenants to keep the Slashdot community from descending into complete anarchy. But without the resources to hire a hundred full-time moderators, Slashdot appeared to be stuck at the same impasse that Mumford had described thirty years before: stay small and preserve the quality of the original community; keep growing and sacrifice everything that had made the community interesting in the first place. Slashdot had reached its “climax stage.”
What did the Commander do? Instead of expanding his pool of special authorized lieutenants, he made everyone a potential lieutenant. He handed over the quality-control job to the community itself. His goals were relatively simple, as outlined in the Frequently Asked Questions document on the site:
1. Promote quality, discourage crap.
2. Make Slashdot as readable as possible for as many people as possible.
3. Do not require a huge amount of time from any single moderator.
4. Do not allow a single moderator a “reign of terror.”
Together, these objectives define the parameters of Slashdot’s ideal state. The question for Malda was how to build a homeostatic system that would naturally push the site toward that state without any single individual being in control. The solution that he arrived at should be immediately recognizable by now: a mix of negative and positive feedback, structured randomness, neighbor interactions, and decentralized control. From a certain angle, Slashdot today resembles an ant colony. From another, it looks like a virtual democracy. Malda himself likens it to jury duty.
Here’s how it works: If you’ve spent more than a few sessions as a registered Slashdot user, the system may on occasion alert you that you have been given moderator status (not unlike a jury summons arriving in your mailbox). As in the legal analogy, moderators only serve for a finite stretch of time, and during that stretch they have the power to rate contributions made by other users, on a scale of -1 to 5. But that power diminishes with use: each moderator is endowed only with a finite number of points that he or she can distribute by rating user contributions. Dole out all your ratings, and your tenure as a moderator comes to an end.
Those ratings coalesce into something that Malda called karma: if your contributions as a user are highly rated by the moderators, you earn karma in the system, giving you special privileges. Your subsequent posts begin life at a higher rating than usual, and you are more likely to be chosen as a moderator in future sessions. This last privilege exemplifies meta-feedback at work, the ratings snake devouring its own tail: moderators rate posts, and those ratings are used to select future moderators. Malda’s system not only encouraged quality in the submissions to the site; it also set up an environment where community leaders could naturally rise to the surface. That elevation was specifically encoded in the software. Accumulating karma on Slashdot was not just a metaphor for winning the implicit trust of the Slashdot community; it was a quantifiable number. Karma had found a home in the database.
Malda’s point system brings to mind the hit points of Dungeons & Dragons and other classics of the role-playing genre. (That the Slashdot crowd was already heavily versed in the role-playing idiom no doubt contributed a great deal to the rating system’s quick adoption.) But Malda had done something more ambitious than simply porting gaming conventions to the community space. He had created a kind of currency, a pricing system for online civics. By ensuring that the points would translate into special privileges, he gave them value. By making one’s moderation powers expendable, he created the crucial property of scarcity. With only one or the other, the currency is valueless; combine the two, and you have a standard for pricing community participation that actually works.
The connection between pricing and feedback is itself more than a metaphor. As a character in Jane Jacobs’s recent Socratic dialogue, The Nature of Economies, observes: “Adam Smith, back in 1775, identified prices of goods and rates of wages as feedback information, although of course he didn’t call it that because the word feedback was not in the vocabulary at the time. But he understood the idea… . In his sober way, Smith was clearly excited about the marvelous form of order he’d discovered, as well he should have been. He was far ahead of naturalists in grasping the principle of negative feedback controls.”
Malda himself claims that neither The Wealth of Nations nor The Dungeon Master’s Guide were heavy in his thoughts in Geek House One. “There wasn’t really
anything specific that inspired me,” Malda says now. “It was mostly trial and error. The real influence was my desire to please users with very different expectations for Slashdot. Some wanted it to be Usenet: anything goes and unruly. Others were busy people who only wanted to read three to four comments a day.” You can see the intelligence and flexibility of the system firsthand: visit the Slashdot site and choose to view all the posts for a given conversation. If the conversation is more than a few hours old, you’ll probably find several hundred entries, with at least half of them the work of cranks and spammers. Such is the fate of any Web site lucky enough to attract thousands of posts an hour.
Set your quality threshold to four or five, however, and something miraculous occurs. The overall volume drops precipitously—sometimes by an order of magnitude—but the dozen or two posts that remain will be as stimulating as anything you’ve read on a traditional content site where the writers and the editors are actually paid to put their words and arguments together. It’s a miracle not so much because the quality is lurking there somewhere in the endless flood of posting. Rather, it’s a miracle because the community has collectively done such an exceptional job at bringing that quality to light. In the digital world, at least, there is life after the climax stage.
*
Slashdot is only the beginning. In the past two years, user ratings have become the kudzu of the Web, draping themselves across pages everywhere you look. Amazon had long included user ratings for all the items in its inventory, but in 1999 it began to let users rate the reviews of other users. An ingenious site called Epinions cultivates product reviews from its audience and grants “trust” points to contributors who earn the community’s respect. The online auction system of eBay utilizes two distinct feedback mechanisms layered on top of each other: the price feedback of the auction bids coupled to the user ratings that evaluate buyers and sellers. One system tracks the value of stuff; the other tracks the value of people.
Emergence Page 14