Everything Is Obvious

by Duncan J. Watts

  This last point presents a problem because when we try to explain the success of the Mona Lisa, it is precisely its attributes on which we focus our attention. If you’re Kenneth Clark, you don’t need to know anything about the circumstances of the Mona Lisa’s rise to fame to know why it happened—everything you need to know is right there in front of you. To oversimplify only slightly, the Mona Lisa is the most famous painting in the world because it is the best, and although it might have taken us a while to figure this out, it was inevitable that we would. And that’s why so many people are puzzled when they first actually set eyes on the Mona Lisa. They’re expecting these intrinsic qualities to be apparent, and they’re not. Of course, most of us, when faced with this moment of dissonance, simply shrug our shoulders and assume that somebody wiser than us has seen things we can’t see. And yet as Sassoon deftly but relentlessly lays out, whatever attributes the experts cite as evidence—the novel painting technique that Leonardo employed to produce so gauzy a finish, the mysterious subject, her enigmatic smile, even da Vinci’s own fame—one can always find numerous other works of art that would seem as good, or even better.

  Of course, one can always get around this problem by pointing out that it’s not any one attribute of the Mona Lisa that makes it so special, but rather the combination of all its attributes—the smile, and the use of light, and the fantastical background, and so on. There’s actually no way to beat this argument, because the Mona Lisa is of course a unique object. No matter how many similar portraits or paintings some pesky skeptic drags out of the dustbin of history, one can always find some difference between them and the one that we all know is the deserving winner. Unfortunately, however, this argument wins only at the cost of eviscerating itself. It sounds as if we’re assessing the quality of a work of art in terms of its attributes, but in fact we’re doing the opposite—deciding first which painting is the best, and only then inferring from its attributes the metrics of quality. Subsequently, we can invoke these metrics to justify the known outcome in a way that seems rational and objective. But the result is circular reasoning. We claim to be saying that the Mona Lisa is the most famous painting in the world because it has attributes X, Y, and Z. But really what we’re saying is that the Mona Lisa is famous because it’s more like the Mona Lisa than anything else.

  CIRCULAR REASONING

  Not everybody appreciates this conclusion. When I explained the argument to an English literature professor at a party once, she practically shouted, “Are you suggesting that Shakespeare might just be a fluke of history?” Well, as a matter of fact, that’s exactly what I was suggesting. Don’t get me wrong: I enjoy Shakespeare as much as any normal person. But I also know that I didn’t acquire my appreciation in a vacuum. Like just about everyone else in the Western world, I spent years of high school laboring over his plays and sonnets. And presumably, like many others, it wasn’t immediately obvious to me what all the fuss was about. Try reading Midsummer Night’s Dream and forget for a moment that you know it’s a work of genius. Right about the point where Titania is fawning over a man with the head of a donkey, you might just start to wonder what on earth Shakespeare was thinking. But I digress. The point is that no matter what my schoolboy brain thought of what it was reading, I was determined to appreciate the genius that my teachers assured us was there. And if I hadn’t, it would have been my failing, not Shakespeare’s—because Shakespeare, like da Vinci, defines genius. As with the Mona Lisa, this outcome may be perfectly justified. Nevertheless, the point remains that locating the source of his genius in the particular attributes of his work invariably leads us in circles: Shakespeare is a genius because he is more like Shakespeare than anyone else.

  Although it is rarely presented as such, this kind of circular reasoning—X succeeded because X had the attributes of X—pervades commonsense explanations for why some things succeed and others fail. For example, an article on the success of the Harry Potter books explained it this way: “A Cinderella plot set in a novel type of boarding school peopled by jolly pupils already has a lot going for it. Add in some easy stereotypes illustrating meanness, gluttony, envy, or black-hearted evil to raise the tension, round off with a sound, unchallenging moral statement about the value of courage, friendship, and the power of love, and there already are some of the important ingredients necessary for a match-winning formula.” In other words, Harry Potter was successful because it had exactly the attributes of Harry Potter, and not something else.

  Likewise, when Facebook first became popular, conventional wisdom held that its success lay in its exclusivity to college students. Yet by 2009, long after Facebook had opened itself to everyone, a report by Nielsen, the ratings company, attributed its success to its broad appeal, along with its “simple design” and “focus on connecting.” Facebook, in other words, was successful because it had exactly the attributes of Facebook, even as the attributes themselves changed completely. Or consider a news story reviewing 2009 movies that inferred from the success of The Hangover that “relatable, non-thinking comedies … are the perfect balm for the recession,” implying in effect that The Hangover succeeded because moviegoers wanted to see a movie like The Hangover, and not something else. In all these cases, we want to believe that X succeeded because it had just the right attributes, but the only attributes we know about are the attributes that X possesses; thus we conclude that these attributes must have been responsible for X’s success.4

  Even when we are not explaining success, we still rely on circular reasoning to make sense of why certain things happen. For example, in another recent news story about an apparent downshift in postrecession consumer behavior, one expert explained the change with the helpful observation that “It’s simply less fun pulling up to the stoplight in a Hummer than it used to be. It’s a change in norms.” People do X, in other words, because X is the norm, and it is normal to follow norms. OK, great, but how do we know that something is a norm? We know because people are following it. And this is no isolated example. Once you start to pay attention, it’s amazing how often explanations contain this circularity. Whether it is women getting the vote, gay and lesbian couples being allowed to marry, or a black man being elected president, we routinely explain social trends in terms of what society “is ready for.” But the only way we know society is ready for something is because it happened. Thus, in effect, all we are really saying is that “X happened because that’s what people wanted; and we know that X is what they wanted because X is what happened.”5

  THE MICRO-MACRO PROBLEM

  The circularity evident in commonsense explanations is important to address because it derives from what is arguably the central intellectual problem of sociology—which sociologists call the micro-macro problem. The problem, in a nutshell, is that the outcomes that sociologists seek to explain are intrinsically “macro” in nature, meaning that they involve large numbers of people. Paintings, books, and celebrities can only be popular or unpopular to the extent that large numbers of people care about them. Firms, markets, governments, and other forms of political and economic organization all require large numbers of people to abide by their rules in order for anything to actually happen. And cultural institutions such as marriage, social norms, and even legal principles have relevance only to the extent that large numbers people believe that they do. At the same time, however, it is necessarily the case that all these outcomes are driven in some way by the “micro” actions of individual humans, who are making the kinds of choices that I discussed in the previous chapter. So how do we get from the micro choices of individuals to the macro phenomena of the social world? Where, in other words, do families, firms, markets, cultures, and societies come from, and why do they exhibit the particular features that they exhibit? This is the micro-macro problem.

  As it turns out, something like the micro-macro problem comes up in every realm of science, often under the label “emergence.” How is it, for example, that one can lump together a collection of atoms and somehow ge
t a molecule? How is it that one can lump together a collection of molecules and somehow get amino acids? How is it that one can lump together a collection of amino acids and other chemicals and somehow get a living cell? How is it that one can lump together a collection of living cells and somehow get complex organs like the brain? And how is it that one can lump together a collection of organs and somehow get a sentient being that wonders about its eternal self? Seen in this light, sociology is merely at the tip of the pyramid of complexity that begins with subatomic particles and ends with global society. And at each level of the pyramid, we have essentially the same problem—how do you get from one “scale” of reality to the next?

  Historically, science has done its best to dodge this question, opting instead for a division of labor across the scales. Physics, therefore, is its own subject with its own set of facts, laws, and regularities, while chemistry is a different subject altogether, with an entirely different set of facts, laws, and regularities, and biology is a whole new ballgame all over again. At some level the laws that apply at different scales must be consistent—one cannot have chemistry that violates the laws of physics—but it is not generally possible to derive the laws that apply at one scale from those that govern the scale below it. Knowing everything about the behavior of individual neurons, for example, would be of little help in understanding human psychology, just as a complete knowledge of particle physics would be of little use in explaining the chemistry of synapses.6

  Increasingly, however, the questions that scientists find most interesting—from the genomics revolution to the preservation of ecosystems to cascading failures in power grids—are forcing them to consider more than one scale at a time, and so to confront the problem of emergence head-on. Individual genes interact with each other in complex chains of activation and suppression to express phenotypic traits that are not reducible to the properties of any one gene. Individual plants and animals interact with each other in complex ways, via prey-predator relations, symbiosis, competition, and cooperation, to produce ecosystem-level properties that cannot be understood in terms of any individual species. And individual power generators and substations interact with each other via high-voltage transmission cables to produce system-level dynamics that cannot be understood in terms of any individual component.

  Social systems are also replete with interactions—between individual people, between individuals and firms, between firms and other firms, between individuals, firms, and markets, and between everyone and the government. Individual people are influenced by what other people are doing or saying or wearing. Firms are affected by what individual consumers want and also by what their competitors are producing, or what their debt holders require of them. Markets are affected by government regulations as well as by the actions of individual firms, and sometimes even of individual people (think Warren Buffett or Ben Bernanke). And governments are swayed by all manner of influences, from corporate lobbyists to opinion polls to stock market indices. In the kinds of systems that sociologists study, in fact, the interactions come in so many forms and carry such consequence, that our own version of emergence—the micro-macro problem—is arguably more complex and intractable than in any other discipline.

  Common sense, however, has a remarkable knack for papering over this complexity. Emergence, remember, is a hard problem precisely because the behavior of the whole cannot easily be related to the behavior of its parts, and in the natural sciences we implicitly acknowledge this difficulty. For example, we do not speak of the genome as if it behaves like a single gene, nor do we speak of brains as if they behave like individual neurons, or ecosystems like individual creatures. That would be ridiculous. When it comes to social phenomena, however, we do speak of “social actors” like families, firms, markets, political parties, demographic segments, and nation-states as if they act in more or less the same way as the individuals that comprise them. Families, that is, “decide” where to go on vacation, firms “choose” between business strategies, and political parties “pursue” legislative agendas. Likewise, advertisers speak of appealing to their “target demographic,” Wall Street traders dissect the sentiment of “the market,” politicians speak about “the will of the people,” and historians describe a revolution as a “fever gripping society.”

  Everyone understands, of course, that corporations and political parties, even families, do not literally have feelings, form beliefs, or imagine the future the way individual people do. Nor are they subject to the same psychological quirks and biases that I discussed in the previous chapter. At some level, we know that the “behavior” of social actors is really a convenient shorthand for the aggregate behavior of large numbers of individuals. Nevertheless, it is so natural to talk this way that the shorthand has become indispensible to our ability to explain things. Imagine trying to recount the history of World War II without talking about the actions of the Allies or the Nazis. Imagine trying to understand the Internet without talking about the behavior of large Internet companies like Microsoft or Yahoo! or Google. Or imagine trying to analyze the debate over healthcare reform in the United States without talking about the interests of Democrats or Republicans, or “special interests.” Margaret Thatcher was famous for having said “There is no such thing as society. There are individual men and women, and there are families.”7 But if we actually tried to apply Thatcher’s doctrine to explaining the world, we wouldn’t even know where to start.

  In social science, Thatcher’s philosophical position goes by the name of methodological individualism, which claims that until one has succeeded in explaining some social phenomenon—the popularity of the Mona Lisa or the relation between interest rates and economic growth—exclusively in terms of the thoughts, actions, and intentions of individual people, one has not fully succeeded in explaining it at all. Explanations that ascribe individual psychological motivations to aggregate entities like firms, markets, and governments might be convenient, but they are not, as the philosopher John Watkins put it, “rock bottom” explanations.8

  Unfortunately, attempts to construct the kind of rock-bottom explanations that methodological individualists imagined have all run smack into the micro-macro problem. In practice, therefore, social scientists invoke what is called a representative agent, a fictitious individual whose decisions stand in for the behavior of the collective. To take a single example, albeit an important one, the economy is composed of many thousands of firms and millions of individuals all making decisions about what to buy, what to sell, and what to invest in. The end result of all this activity is what economists call the business cycle—in effect, a time series of aggregate economic activity that seems to exhibit periodic ups and downs. Understanding the dynamics of the business cycle is one of the central problems of macroeconomics, in no small part because it affects how policy makers deal with events like recessions. Yet the mathematical models that economists rely on do not attempt to represent the vast complexity of the economy at all. Rather, they specify a single “representative firm” and ask how that firm would rationally allocate its resources given certain information about the rest of the economy. Roughly speaking, the response of that firm is then interpreted as the response of the economy as a whole.9

  By ignoring the interactions between thousands or millions of individual actors, the representative agent simplifies the analysis of business cycles enormously. It assumes, in effect, that as long as economists have a good model of how individuals behave, they effectively have a good model for how the economy behaves as well. In eliminating the complexity, however, the representative-agent approach effectively ignores the crux of the micro-macro problem—the very core of what makes macroeconomic phenomena “macro” in the first place. It was for precisely this reason, in fact, that the economist Joseph Schumpeter, who is often regarded as the founding father of methodological individualism, attacked the representative-agent approach as flawed and misleading.10

  In practice, however, methodological individualists have l
ost the battle, and not just in economics. Pick up any work of history, sociology, or political science that deals with “macro” phenomena, like class, race, business, war, wealth, innovation, politics, law, or government, and you will find a world populated with representative agents. So common is their use in social science, in fact, that the substitution of a fictitious individual for what is in reality a collective typically happens without so much as an acknowledgment, like the magician placing the rabbit in the hat while the audience is looking elsewhere. No matter how it is done, however, the representative agent is only and always a convenient fiction. And no matter how we try to dress them up in mathematics or other finery, explanations that invoke representative agents are making essentially the same error as commonsense explanations that talk about firms, markets, and societies in the same terms that we use to describe individual people.11

  GRANOVETTER’S RIOT MODEL

  The sociologist Mark Granovetter highlighted this problem using a very simple mathematical model of a crowd poised on the brink of a riot. Say a crowd of a hundred students is gathered in a town square, protesting the government’s proposed increase in student fees. The students are angry about the new policy and frustrated with their lack of input to the political process. There’s a possibility of things getting out of hand. But being educated, civilized people, they also understand that reason and dialogue are preferable to violence. To oversimplify somewhat, each individual in the crowd is torn between two instincts—one to go berserk and smash things up, and the other to remain calm and protest peacefully. Everyone, whether they are conscious of it or not, has to make a choice between these two actions. But they are not making a choice between violence and peaceful protest independently—they are doing so, at least in part, in response to what other people are doing. The greater the number of individuals who engage in a riot, the more likely their efforts will force the politicians to pay attention, and the less likely that any one of them will be caught and punished. Also, riots have a primal energy of their own that can undermine otherwise strong social conventions against physical destruction, even skewing our psychological estimation of risk. In a riot, even sensible people can go crazy. For all these reasons, the choice about whether to remain calm or to engage in violence is subject to the general rule that the more other people are rioting, the more likely any particular individual is to join in.