What the Music Lab experiment really showed, therefore, was remarkably similar to the basic insight from Granovetter’s riot model—that when individuals are influenced by what other people are doing, similar groups of people can end up behaving in very different ways. This may not sound like a big deal, but it fundamentally undermines the kind of commonsense explanations that we offer for why some things succeed and others fail, why social norms dictate that we do some things and not others, or even why we believe what we believe. Commonsense explanations sidestep the whole problem of how individual choices aggregate to collective behavior simply by replacing the collective with a representative individual. And because we think we know why individual people do what they do, as soon as we know what happened, we can always claim that it was what this fictitious individual—“the people,” “the market,” whatever—wanted.
By pulling apart the micro-macro problem, experiments like Music Lab expose the fallacy that arises from this form of circular reasoning. Just as you can know everything about the behavior of individual neurons and still be mystified by the emergence of consciousness in the human brain, so too you could know everything about individuals in a given population—their likes, dislikes, experiences, attitudes, beliefs, hopes, and dreams—and still not be able to predict much about their collective behavior. To explain the outcome of some social process in terms of the preferences of some fictitious representative individual therefore greatly exaggerates our ability to isolate cause and effect.
For example, if you’d asked the 500 million people who currently belong to Facebook back in 2004 whether or not they wanted to post profiles of themselves online and share updates with hundreds of friends and acquaintances about their everyday goings-on, many of them would have likely said no, and they probably would have meant it. The world, in other words, wasn’t sitting around waiting for someone to invent Facebook so that we could all join it. Rather, a few people joined it for whatever reasons and began to play around with it. Only then, because of what those people experienced through using the service as it existed back then—and even more so because of the experiences they created for one another in the course of using it—did other people began to join. And then other people joined because those people joined, and so on, until here we are today. Yet now that Facebook is tremendously popular, it just seems obvious that it must have been what people wanted—otherwise, why would they be using it?
This is not to say that Facebook, the company, hasn’t made a lot of smart moves over the years, or doesn’t deserve to be as successful as it is. Rather, the point is just that the explanations we give for its success are less relevant than they seem. Facebook, that is, has a particular set of features, just as Harry Potter and the Mona Lisa have their own particular sets of features, and all of them have experienced their own particular outcomes. But it does not follow that those features caused the outcomes in any meaningful way. Ultimately, in fact, it may simply not be possible to say why the Mona Lisa is the most famous painting in the world or why the Harry Potter books sold more than 350 million copies within ten years, or why Facebook has attracted more than 500 million users. In the end, the only honest explanation may be the one given by the publisher of Lynne Truss’s surprise bestseller, Eats, Shoots and Leaves, who, when asked to explain its success, replied that “it sold well because lots of people bought it.”
It may not surprise you to learn that many people do not particularly like this conclusion. Most of us would be prepared to admit that our decisions are influenced by what other people think—sometimes, anyway. But it’s one thing to acknowledge that once in a while our behavior gets nudged this way or that by what other people are doing, and it’s quite another to concede that as a consequence, true explanations for the success of an author or a company, unexpected changes in social norms, or the sudden collapse of a seemingly impregnable political regime may simply lie beyond our reach. When faced with the prospect that some outcome of interest cannot be explained in terms of special attributes or conditions, therefore, a common fallback is to assume that it was instead determined by a small number of important or influential people. So it is to this topic that we turn next.
CHAPTER 4
Special People
It may seem hard to believe in a time when “social networking” has become so commonplace an idea that it shows up in everything from feature films to Foster’s beer commercials, but it wasn’t that long ago—as recently as the mid-1990s—that the study of social networks was relatively obscure, pursued mostly by a small cadre of mathematically inclined sociologists interested in mapping the social interactions among individuals.1 The field has exploded in recent years, in large part because fast computers, along with communication technologies like e-mail, cell phones, and social networking sites such as Facebook have made it possible to record and analyze these interactions with great precision, even for hundreds of millions of people at a time. Nowadays, thousands of computer scientists, physicists, mathematicians, and even biologists count themselves as “network scientists,” and new discoveries about the structure and dynamics of networked systems arrive daily.2
SIX DEGREES OF SEPARATION
Back in 1995, however, when I was a graduate student at Cornell studying the synchronization of chirping crickets, all this was in the future. Back then, in fact, the idea that everyone in the world is connected through a giant social network through which information, ideas, and influence might flow, was still sufficiently novel that when my father asked me during one of our regular phone conversations if I’d ever heard of the notion that “everyone in the world is only six handshakes away from the president of the United States,” I assumed it was folklore. And in some sense it was. People have been fascinated with what sociologists call the small-world problem for nearly a century, since the Hungarian poet Frigyes Karinthy published a short story called “Chains” in which his protagonist boasts that he can connect himself to any other person in the world, whether a Nobel Prize winner or a worker in a Ford Motor factory, through a chain of no more than five acquaintances. Four decades later, in her polemic on urban planning The Death and Life of Great American Cities, the journalist Jane Jacobs described a similar game, called messages, that she used to play with her sister when they first moved to New York:
The idea was to pick two wildly dissimilar individuals—say a headhunter in the Solomon Islands and a cobbler in Rock Island, Illinois—and assume that one had to get a message to the other by word of mouth; then we would each silently figure out a plausible, or at least possible, chain of persons through whom the message could go. The one who could make the shortest plausible chain of messages won.
But how long are these chains in reality? One way to answer the question would be to map out all the links in the social network of the whole world and then simply count by brute force how many people you can reach in one “degree of separation,” how many at “two degrees,” and so on, until you have reached everyone. In Jacobs’s day that was impossible, but in 2008 two computer scientists at Microsoft Research got somewhat close when they computed the length of paths connecting pairs of individuals in Microsoft’s 240-million-strong instant messenger network, where being “friends” in this case meant being on each other’s buddy lists.3 On average they found that people were separated by about seven steps—remarkably close to the six handshakes that my father had mentioned. Yet this can’t be the real answer to the question. The characters in Jacobs’s fictional game didn’t have access to this network, so they couldn’t have computed the paths the way the Microsoft researchers did even if they had the computing power to do so. Clearly they must have used some other method to direct their messages. And indeed, according to Jacobs, they did:
The headhunter would speak to the headman of his village, who would speak to the trader who came to buy copra, who would speak to the Australian patrol officer when he came through, who would tell the man who was next slated to go to Melbourne on leave, etc. Down at the other end,
the cobbler would hear from his priest, who got it from the mayor, who got it from the state senator, who got it from the governor, etc. We soon had these close-to-home messengers down to a routine for almost everybody we could conjure up, but we would get tangled up in long chains at the middle until we began employing Mrs. Roosevelt. Mrs. Roosevelt made it suddenly possible to skip whole chains of intermediate connections. She knew the most unlikely people. The world shrank remarkably.4
Jacobs’s solution, in other words, assumes that social networks are organized in a hierarchy: Messages flow up the hierarchy from the periphery and then back down again, with high-status figures like Mrs. Roosevelt occupying the critical center. We are so used to a world of hierarchies—whether inside formal organizations, across the economy, or in society—that it is natural to assume that social networks should be hierarchical as well. Karinthy, in fact, used a similar line of reasoning to Jacobs’s, where in place of Mrs. Roosevelt he invoked Mr. Ford, writing that “to find a chain of contacts linking myself with an anonymous riveter at the Ford Motor Company … The worker knows his foreman, who knows Mr. Ford himself, who in turn is on good terms with the director general of the Hearst publishing empire. It would take but one word from my friend to send a cable to the general director of Hearst asking him to contact Ford who could in turn contact the foreman, who could then contact the riveter, who could then assemble a new automobile for me, should I need one.”
As plausible as this method sounds, however, it is not how messages actually propagate through social networks, as we know now from a series of “small-world experiments” that began not long after Jacobs was writing. The first of these experiments was conducted by none other than Stanley Milgram, the social psychologist whose subway experiment I discussed in Chapter 1. Milgram recruited three hundred people, two hundred from Omaha, Nebraska, and the other hundred from around Boston, to play a version of the messages game with a Boston stockbroker who was a friend of Milgram’s and who had volunteered to serve as the “target” of the exercise. Much as in Jacobs’s imaginary version, participants in Milgram’s experiment knew whom they were trying to reach, but could only send the message to someone whom they knew on a first-name basis; thus each of the three hundred “starters” would send it to a friend, who would send it to a friend, and so on, until someone either refused to participate or else the message chain reached the target. As luck would have it, sixty-four of the initial chains made it all the way to their destination, and the average length of those that did was indeed about six; hence the famous phrase “six degrees of separation.”5
But although Milgram’s subjects were able to find paths as short as those hypothesized by Karinthy and Jacobs, it wasn’t because they employed Mrs. Roosevelt or anyone like her. Instead, ordinary people passed messages to other ordinary people, tracking along the same social stratum rather than going up and down the hierarchy as both Karinthy and Jacobs imagined. Nor did the chains get tangled up in the middle as Jacobs worried they might. Instead they experienced their greatest difficulties after they had already gotten close to their targets. Social networking, it seems, is less like a pyramid than it is like a game of golf—where, the old adage goes, you “drive for show, putt for dough.” When you are far away from the target, that is, it’s relatively easy to jump large distances simply by sending the message to someone in the right country, and from there to someone in the right city, and then to someone in the right profession. But once you get close to the target, big jumps don’t help you anymore and messages have a tendency to bounce around until they find someone who knows the target.
Nevertheless, Milgram still found that not all message handlers are created equal. In fact, of the sixty-four messages that got through, nearly half of them were delivered to the target by one of three people, and half of those—sixteen chains—were delivered by a single person, “Mr. Jacobs,” a clothing merchant who was a neighbor of the target. Struck by this concentration of messages into the hands of a few individuals, Milgram speculated that what he called sociometric stars might be important to understanding how the small-world phenomenon worked.6 Milgram himself didn’t conclude much more than that, but three decades later, in an essay called “Six Degrees of Lois Weisberg,” New Yorker writer Malcolm Gladwell used Milgram’s finding about Mr. Jacobs to make the argument that “a very small number of people [like Mrs. Weisberg] are linked to everyone else in a few steps, and the rest of us are linked to the world through those few.”7 In other words, even though Mr. Jacobs and Ms. Weisberg are not “important” in the same way that Mrs. Roosevelt or Mr. Ford were important, from a network perspective they end up serving the same kind of function—like hubs in an airline network that we necessarily pass through in order to get from one part of the world to another.
Like Jacobs’s hierarchy, the airline network metaphor is an appealing one, but it says more about how we would organize the world if given the opportunity to do so than it says about how the world is actually organized. If you think about it for minute, in fact, the metaphor is actually quite implausible. Some people clearly have more friends than others. But people are not like airports—they can’t just tack on an extra wing when they need to handle more traffic. As a result, the number of friends that people have doesn’t vary by nearly as much as the traffic in airports. An average person, for example, has a few hundred friends, while the most gregarious top out around a couple of thousand—roughly ten times as many. That is a big difference, but not remotely comparable to a true hub like O’Hare, which handles thousands of times as many passengers as a small airport. So how is it that connectors in social networks can nevertheless act like hubs in airline networks?8
In fact, they do not, as my collaborators Roby Muhamad and Peter Dodds and I found several years ago when we replicated Milgram’s original experiment—only this time we used e-mail in place of physical packets, allowing us to work on a much larger scale. Whereas Milgram had three hundred initial senders in two cities attempting to reach a single target in Boston, we had more than twenty thousand chains in search of one of eighteen targets in thirteen different countries around the world. By the time the experiment had ended, the chains had passed through over 60,000 people in 166 countries. Using some more up-to-date statistical analysis than Milgram had available to him, we were also able to estimate not only the length of the chains that made it to their targets, but also how long the chains that failed would have been had they continued. Our main finding was remarkably close to Milgram’s—roughly half of all chains should be expected to reach their targets in seven steps or fewer. Given the differences between the two experiments, which were conducted on very different scales using different technologies and nearly forty years apart, it is actually sort of amazing that the results accorded so closely, and provides strong support for the claim that many people—although certainly not all people—can connect to one another through short chains.9
Unlike Milgram’s findings, however, we discovered no “hubs” in the delivery process. Rather, messages reached their targets through almost as many recipients as there were chains. We also asked people why they chose the next person in the chain, and here, too, we discovered little evidence of hubs or stars. Subjects in small-world experiments, it turns out, do not typically pass messages to their highest-status or most-connected friends. Instead, they pass them to people they think have something in common with the target, like geographic proximity or a similar occupation, or they simply pass them to people they think will be likely to continue passing it along. Ordinary individuals, in other words, are just as capable of spanning critical divides between social and professional circles, between different nations, or between different neighborhoods, as exceptional people. When you want to get a message to a graduate student in Novosibirsk, Russia, for example, you don’t think about whom you know who has a lot of friends, or goes to lots of parties, or has connections to the White House. You think about whether you know any Russians. And if you don’t know any Russians, then
maybe you know someone from Eastern Europe, or someone who has traveled to Eastern Europe, or has studied Russian, or who lives in a part of your city that is known for its Eastern European immigrants. Mrs. Roosevelt, or Lois Weisberg for that matter, may indeed connect many people. But those same people have many other ways of connecting as well. And it is these other, less obvious ways that they tend to actually use, if only because there are so many more of them.
The overall message here is that real social networks are connected in more complex and more egalitarian ways than Jacobs or even Milgram imagined—a result that has now been confirmed with many experiments, empirical studies, and theoretical models.10 In spite of all this evidence, however, when we think about how social networks work, we continue to be drawn to the idea that certain “special people,” whether famous wives of presidents or gregarious local businessmen, are disproportionately responsible for connecting the rest of us. Evidence, in fact, seems to have very little to do with why we think this way. After all, Jacobs was writing years before Milgram’s experiments and long before anyone had the kind of data that might have supported her claim about Mrs. Roosevelt. So wherever she got the idea from, it obviously wasn’t based on any actual evidence. Rather, it seems that Jacobs was drawn to the idea that a few special people connect everyone else simply because without invoking such people it’s hard to come with any explanation at all. The result is that no matter how many times the evidence rules out one kind of special person, we simply insert another. If it’s not Mrs. Roosevelt, then it must be Lois Weisberg, and if it’s not Lois Weisberg, then it must be Mr. Jacobs the clothing merchant. And if it’s not him, then it must be our friend Ed who seems to know everyone. “It’s got to be someone special.” We feel compelled to conclude: “How else could it work?”
Everything Is Obvious Page 9