Nature, nurture, or network?

Illustration by Mark Zurolo ’01MFA and Celia Poirier

This network of close friendships in a college dorm, from Christakis and Fowler's book Connected, illustrates some of the basic qualities of different positions in a network. Each dot, or node, represents a person and each line a friendship. Nodes A, B, and D are all on the periphery; Node C is central. But Node A differs from B and D in that, for instance, Node A's friends tend also to be friends with each other. These differing relationships all have different influences on individual behavior. View full image

One day in the late 1990s, Nicholas Christakis ’84 was on the South Side of Chicago visiting a woman who had Alzheimer’s disease. Christakis was then a young physician and social scientist at the University of Chicago, taking care of terminally ill patients and also studying the widower effect—in which the death of one spouse dramatically worsens the likelihood of death for the other.

That day’s patient was gradually dwindling away, attended by her daughter. “The daughter was exhausted from caring for her mother,” Christakis recalls. The daughter’s husband was also sick from his wife’s exhaustion. Then Christakis got a phone call from one of the husband’s pals, “depressed by what was happening to his friend.”

It dawned on Christakis that the widower effect was not just about husbands and wives, or even pairs of people like the mother and her daughter. It rippled outward across networks of family, friends, and coworkers. Most surprisingly, given the narrow focus of his own work up to that point, it wasn’t just about death. “So I started to see the world in a completely new way,” Christakis recalled, in a 2010 TED talk, “and I became obsessed with how it might be that we’re embedded in these social networks, and how they affect our lives.”

Questions about the nature of networks have dominated his research ever since, first at Chicago, then during a 12-year stint at Harvard, where his growing interest in networks and biosocial science ultimately led him to give up his medical practice, and now at Yale, to which Christakis returned in summer 2013 as a professor of both sociology and medicine. (His full title is the Sol Goldman Family Professor of Social and Natural Science.)

Over the past few years, his work illuminating the nature of social networks has won Christakis recognition not just within the scholarly world, but on Time magazine’s 2009 list of 100 “people who affect the world,” and on Foreign Policy’s 2009 and 2010 lists of top global thinkers. His 2009 book, Connected: How Your Friends’ Friends’ Friends Affect Everything You Feel, Think, and Do, has appeared in nearly 20 languages. He wrote it, appropriately, with his best friend and longtime research collaborator James H. Fowler ’97MA, now at the University of California, San Diego. Their work demonstrating the contagious nature of everything from obesity to altruism has stirred up considerable debate in the research world. It has also suggested powerful new ways to intervene in networks—for instance, to speed the switch to generic drugs, or to slow the spread of sexually transmitted diseases.

Mark Ostow

Networks, Christakis believes, are structured basically the same way across cultures and epochs. View full image

What Christakis and Fowler are proposing amounts to a strikingly different way of looking at our own lives, adding a new “n’ to the familiar dichotomy of nurture and nature: we are also creatures of our social networks—simultaneously individuals and, says Christakis, intimately connected parts of a superorganism. We are metagenomic. It’s a term that has lately come into common usage to describe how the microbial genomes in and around our bodies help shape our physical and emotional well-being. But Christakis means it in a broader sense, too. Each of us also “lives in the sea of genes of others, others with whom we have chosen to connect.” Our friends’ genes, for diverse traits, may help determine how our own genes are expressed and thus who we are.

Christakis believes, moreover, that three recent developments make it possible for the first time to understand how these networks function: first, cell phones, Twitter feeds, medical administrative records, and countless other sources now make possible “massive, passive” gathering of data about social networks. New computational methods also allow researchers to identify social patterns in this sea of data and begin to make sense of them. And finally, inexpensive and widely available DNA sequencing technologies provide a window into the genetic character of these networks.

In another 2010 TED talk, Christakis likened the combined effect to the moment “when Galileo came to use the telescope and could see the heavens in a new way, or Leeuwenhoek invented the microscope and could see biology in a new way.” What’s opening up this time is the hidden universe of our own social behavior.

“Maybe intelligence isn’t about making tools, but about making friends,” Christakis tells his audience, at one of the get-acquainted talks he has been giving around campus. “Our need for social networks may have shaped our ability to think.” He is a contagiously enthusiastic speaker, both erudite and animated, his hands constantly moving, palms up to suggest a possibility, than waggling for an approximation. He has a tendency to talk much too fast, because there is so much to say, but he also says it with a certain playfulness.

When he puts up an animated graphic of shifting social relationships over 30 years among test subjects in the Framingham Heart Study, he remarks that it took five years to make. Then he adds that his teenage daughter, Lena, has described it as “more expensive than Avatar and much less interesting.” For him, though, these maps have the fascination of living organisms.

On the map of a social network, a dot stands in for each person. (See illustration above) The dots, or nodes, are strewn in a seemingly random pattern across a two-dimensional playing surface. Lines, representing friendships, link each node to four or five others, forming a little group of friends. Some groups overlap in a dense cluster at the middle of the network. Others drift out on the periphery, tethered to the larger network by only one or two threads. Centrality, Christakis tells his audience, is a key to understanding how networks function.

Being at the center of a group (like Node C) might seem better—for instance, if you’re hoping to pick up your friends’ gossip. But being on the outskirts (like Node D) can have its advantages, if what’s going around instead is a deadly virus. Advantages and disadvantages also apply to other aspects of a node’s position, he says. “If the challenge is to kill a mastodon,” Node A can band together quickly with his friends, because the people in his tight-knit group all know each other. But “if the job is to find the mastodon,” then it’s better to be Node B, whose friends have connections with other groups. Node B can ask his friends, who can ask their friends, “Did you see the mastodon? Did you see the mastodon?”

The hunter-gatherer reference is deliberate. Christakis believes that the basic structure of social networks has been with us as long as we have been human. Moreover, it hasn’t changed much despite the invention of agriculture, cities, and telecommunications. “If you talked with your great-grandmother” who had no phone or “my teenage daughter who has a phone in her pocket,” he says, they’d both have the same small circle averaging 4.5 close friends.

So would the Hadza, a group of fewer than a thousand people who still follow the traditional hunter-gatherer way of life in north-central Tanzania. Christakis and a postdoc, Coren Apicella, now at the University of Pennsylvania, recently set out to construct a kind of “Hadza Facebook,” on the theory that the Hadza are the closest living approximation of how our ancestors lived before agriculture. (With any other researcher, Apicella says, “I would’ve walked out thinking, ‘This is impossible.’” But Christakis’s enthusiasm—“This is dynamite! We have to study the social networks of the Hadza!”—made it seem almost reasonable.)

“Humans are unusual as a species,” the resulting study in Nature began, “in the extent to which they form long-standing, nonreproductive unions with unrelated individuals—that is, we have friends.” Christakis and Apicella went on to argue that all the same basic network structures operate whether among the Hadza or, say, undergraduates at Harvard—the same variation in the number of social ties (degree distribution), the same likelihood that two of a person’s friends will in turn be friends with each other (transitivity), the same odds that an outbound social tie will elicit an inbound tie from the same person (reciprocity), the same inclination of similar people to form ties together (homophily), and even the same clique-ish knack of popular people for befriending other popular people (degree assortativity).

Christakis’s decision to return to Yale was also a network phenomenon. It started a few years ago when he gave a talk at the University of Pennsylvania and an old colleague asked him to consider coming back to Philadelphia. Christakis had done his residency in internal medicine and earned a PhD in sociology there, after having taken his medical degree and a master’s in public health at Harvard. (Academic cross-training was, and remains, his thing.) His wife, Erika, a childhood educator and public health advocate, had also earned a degree from the University of Pennsylvania. But a Yale alumnus passed along word of the Pennsylvania bid to Peter Salovey ’86PhD, then Yale’s provost. On the resulting visit to New Haven, Christakis says, he “fell in love with the place.”

Or back in love: Christakis was born in New Haven in 1962, when both his parents were Yale graduate students. The family returned to Greece when he was three, and Greek became his first language. (Greek mythological references still turn up routinely in his speech: he says, for instance, that Yale appealed to him this time because it seemed like a place where new ideas could arise from the cross-fertilization of disciplines, “like Daedalus being inspired to invent the first saw after seeing a fish’s jawbone on the beach.”) His father Alexander’s work as a consultant and futurist brought the family back to this country when Nicholas was six. His mother Eleni became a high school science teacher and then a clinical psychologist in Washington, DC. When it was time to choose a college, Christakis didn’t want to go any place but Yale. He went on to graduate with the top prize for science in his class. He was, he says, “incredibly nerdy.”

The off-note in this otherwise ebullient history is that much of the trajectory of his career after Yale derived from a childhood obsession with death. Both his clinical work and his research focused on hospice care and on finding ways to improve medical decisions around the end of life. His doctoral thesis became his first book, Death Foretold: Prophecy and Prognosis in Medical Care. At one point, his thesis adviser worried aloud about this focus: “Do you think it might have something to do with your childhood?” Christakis had immersed himself so thoroughly, so nerdily, in his work that he immediately replied, “It never occurred to me.”

In fact, his career choice had everything to do with the terminal cancer with which his mother Eleni struggled throughout his childhood. “As a boy,” he says, “all I wanted to know was would my mother live or die.” That drove him to “prognostication and death, and how doctors make predictions.” By the time she finally died, he was 25 and in medical school. The career choice was another network phenomenon: his two brothers, Dimitri Christakis, Yale 1986, and Quan-Yang Duh, Yale 1977, also became physicians.

For Nicholas, though, the switch to networks and social science made a better fit. The scholarly intensity aside, he is a highly social, outgoing figure, or, as an interviewer for the Harvard Crimson phrased it, “a total baller.” A college roommate, the author Hampton Sides ’84, recalls that Christakis would occasionally arise from his books, “put on a lot of cologne, the dancing clothes with the bling, and become very Greek” with “a big Mediterranean fro” like “a Saturday Night Fever Adonis.” He was also an aficionado of good Scotch. (He and Fowler still make bets with bottles of Oban over key research questions like whether Modern Apizza or Pepe’s is better.)

At Harvard, he and Erika served as co-masters of a residential house, while their three children were in school, and opened their home as a refuge for the community. When the same Crimson interviewer asked Christakis what he liked best about the job, he replied, “You get to dress up in ridiculous costumes.” (His polar bear outfit was apparently a big deal.) Erika Christakis has also joined the Yale faculty, as a lecturer at the Edward Zigler Center in Child Development and Social Policy. They announced the move to their son Lysander, according to Nicholas, soon after he took early admission to Yale, saying, “Good news, Lysander, mommy and daddy are coming to college with you.”

The network research by Christakis and Fowler first attracted public attention in 2007 when they reported that the “obesity epidemic” wasn’t just a phrase, but an apt description of how weight gain spreads through social networks, even across relatively distant connections. Or as they write in Connected: “Your friend’s husband’s coworker can make you fat.” One particularly hostile critic accused them of “chronic widespread misuse of statistics.” Another characterized the general scholarly response as “respecticism,”—that is, a mixture of respect with a skeptical expectation of further evidence. Andrew C. Thomas, a statistics and social networks scholar at Carnegie Mellon University, reanalyzed their data and argues that their results “were an artifact of their study design and not evidence of contagion.” The issue was partly that “this is very new math,” Thomas says, and partly that the Framingham Heart dataset was inadequate for their purposes. “There would be hope with a much more fine-grained data set in the future that they will be able to show that those effects might be real.” But such a data set would have cost $25 million to assemble, “so they did the best they could with what they had.”

Christakis cites Teddy Roosevelt’s remark that it’s not the critics who count, but the person who acts, “the man in the arena,” who may fall short, “because there is no effort without error and shortcoming” but who nonetheless strives for “the triumph of high achievement.” With his characteristic knack for being lofty and down in the dirt, more or less simultaneously, he adds that critics “sometimes piss me off, and sometimes piss him off,” meaning Fowler. “When they piss us both off, we strike.” The two have gone on to provide supporting evidence not just for the contagious nature of obesity, but for drinking, tobacco use, happiness, sleep, exercise, the prescribing behavior of physicians, and a variety of other phenomena that, Christakis says, have been confirmed by other labs. They’ve also refined their methodology, weeded out potential confounding factors, and branched out into new forms of online experimentation. Still, one observer, Columbia University’s Andrew Gelman, says, “I think they’ve gradually backed down from their strong claims, toward a more general sense that their results are interesting and suggestive.”

Illustration by Mark Zurolo ’01MFA and Celia Poirier

This ring network maps the sexual relationships among half the students in a Midwestern high school over a period of 18 months. (Orange nodes represent females, blue nodes males.) Node X represents a student who had an STD. Node A and Node B are each five steps from X—but Node A is more vulnerable, because the STD can travel in either direction around the ring to reach him. View full image

The concern Christakis and Fowler themselves frequently raise about their research is the “So what?” question. That is, the insights into social networks are intriguing, but what can you do with them?

A lot, they say. In Connected, they present one startling map of a social network of romantic pairings over 18 months in a Midwestern high school. The network included person X, who had an STD. The map shows how person A was more vulnerable than person B to the STD, even though each of them was five steps from X and each had had unprotected sex with just three partners. Their personal risk wasn’t only about individual behavior. It was also about the structure of the network and their dramatically different places within it.

But since nobody inside a network has the perspective to see where they fit in it, so what? Christakis and Fowler argue that public health workers could do better at blocking the spread of a disease if they collected network data and tailored their approach accordingly. In the network at that high school, for instance, there wasn’t a lot of “transitivity”—that is, overlap or duplication among sexual partners. Students largely obeyed the “no-partner-swapping rule”: when a boy and a girl dumped their current partners to become a couple, the dumped parties did not become a couple in turn. With most people being connected to the larger network by only a single link, a school-wide awareness program was the likeliest way to stop the STD outbreak, according to Christakis and Fowler. But in other types of networks, it would be quicker to target a core of individuals who are highly sexually active.

It is, however, often impractical to map any human network, much less a sexual one, meaning again: So what? Christakis says that it’s possible to tweak network behaviors without necessarily mapping them. For instance, the “friendship paradox” says that if you ask each person in a group to name a friend, that named friend will tend to have more friends than the person you are asking. At a cocktail party, say, the suggested names will gravitate toward the host, not the wallflowers. It’s a shortcut for getting at the center of the social network.

Christakis and Fowler tested the idea during a 2009 H1N1 outbreak at Harvard. They found that the more-central individuals got flu two weeks before the random individuals who named them—and as much as 46 days before the epidemic peaked. With current methods, says Christakis, epidemiologists generally spot an outbreak weeks after it occurs. Using network methods instead would allow them to predict and perhaps preempt the flu epidemics that now kill 49,000 Americans every year. Where epidemiologists must now immunize 95 percent of the population to be effective against a highly contagious disease like measles, they could in theory reach the same level of protection immunizing just 30 percent.

When Christakis outlines these ideas during a talk to faculty at the Yale medical school, his audience soon becomes… let’s say reskepticited, a mind-blown blend of respectful, skeptical, and unmistakably excited. When he tells them that identifying central individuals on Twitter makes it possible to predict what’s going to go viral nine days from now, someone inquires, “What about stocks?” (They are doctors, after all.) When he talks about how network strategies can get people to adopt beneficial behaviors more quickly, someone else suggests, “Could you try this on Congress?” With “massive, passive” gathering of data on American citizens by the National Security Agency in the news, a third person wants to know, “how do you keep network ideas and technology from getting into the wrong hands?”

“I know it’s getting in the wrong hands,” Christakis replies. Marketers, hedge funds, and almost every other organization imaginable are all looking for the next advantageous angle, and the new field of computational social science is booming as a result. It is the tantalizing nature of such research: the same network methods that can accelerate the switch to a lower-cost generic drug can serve equally well to increase use of its more profitable brand-name counterpart. The same tools that prove highly effective for getting out the vote can become tricks to keep certain voters at home. One of the functions of networks is to magnify small effects, more or less agnostically: they can spread love. Or they can spread Ebola.

Thinking about how his findings might be misused, Christakis recounts the story of one of his Yale professors, the botanist and bioethicist Arthur Galston. Galston’s doctoral research focused on finding a chemical means to make soybeans flower earlier. Decades after that research, he was devastated to learn that his discoveries had become the basis for the herbicide Agent Orange and that its misuse by the US military as a defoliant in Vietnam had killed, maimed, or permanently disabled hundreds of thousands of people.

But Christakis is in fact far more optimistic than this story suggests. “I believe that the work we are doing will make the world more resistant to disease,” he says, “will make the world more cooperative, will make the world more creative and more civic-minded, and will improve the quality of our society.”

This thought, this dream, makes him reach for a 1968 Robert F. Kennedy speech about the flawed concept of Gross National Product, and in his enthusiasm he reads an excerpt aloud: GNP, Kennedy said, “measures neither our wit nor our courage, neither our wisdom nor our learning, neither our compassion nor our devotion to our country; it measures everything, in short,” and there is an unmistakable hitch in Christakis’s voice, “except that which makes life worthwhile. And it can tell us everything about America except why we are proud that we are Americans.”

The emotion in his voice seems, at that moment, to define him. Christakis is that rare thing in our cynical era—a man still willing to reach for a grand, romantic ideal. He thinks he can tell us something about who we are not just as Americans, but as humans. He understands that he cannot know, any more than Arthur Galston, who will use his research, and that of others in his field, or to what end. But he believes, even so, that he can make it work to build a better world.

It will be interesting to watch him try.