Does the secret to social networking lie in the remote jungle?
The highlands of Honduras’s Copán region, on the country’s Western border with Guatemala, remain nearly as socially isolated today as when the Mayans built one of their greatest civilizations there thousands of years ago. Far from Facebook, Instagram, or Twitter, however, researchers here are studying social networks, trying to map the true extent to which one’s connectivity can influence behavior and, more importantly, be leveraged to achieve positive outcomes.
The $4 million project, backed by the Gates Foundation, is being led by political scientist James Fowler and physician-sociologist Nicholas Christakis through Yale’s Human Nature Lab. Christakis and Fowler have long been at the vanguard of network science, the idea that one’s social circles can have a ripple effect on behavior, good and bad. In Honduras, the pair hope to overcome one of the most persistent challenges to better global health: Introducing best practices and resources, whether relating to clean water, personal health, or disease, is of little value unless they are systematically adopted by the target community.
Yet the Honduras datasets will serve an entirely different purpose as well — as an important critique of contemporary social science, which Christakis and Fowler believe is often not relevant in the real world nor used frequently enough to solve pressing global problems. Indeed, the two research partners are fed up with how little of the academic social network research has been put to practical uses, particularly in the developing world. They feel an urgency to push the emerging field of social network research in a new direction. “There’s a policy side and ethical side to network science that needs attention,” Christakis says. “For the last 100 years, we have turned to physical and biological sciences for human welfare. In the 21st century, social sciences offer deeper promise for human social welfare.”
After years of observational studies and lab experiments, Christakis and Fowler are hoping to find answers in Copán.
Researchers have long wondered how to tease out causality in social networks and grapple with the question of what’s called “homophily,” or the “birds of a feather” tendency for similar people to group together. Do we primarily choose friends who are like us, or do we just become more like one another through peer influence? In their past work, Christakis and Fowler have struggled to convincingly address the issue of homophily, leading some critics to suggest their findings so far have been interesting correlations rather than causations.
Set to run through at least 2018, the Honduras experiment is, in many ways, an answer to that criticism. The randomized controlled trial has been designed to precisely track behavioral changes — as well as cause and effect.
A team of 100 logistics and research personnel has been organized to map the social networks of up to 40,000 people across 160 villages throughout Copán. Dozens of students and researchers this summer trooped through the mountainous terrain to take a census of the villages and get consent from the subjects. With the aid of computer software, the networks will eventually be mapped by asking villagers to point out friends’ faces among photos displayed on tablet screens.
Once these network maps are complete, the researchers will introduce new infant-care regimens. The goal is to find a shortcut to influencing human behavior in a developing world setting, or as Christakis puts it, creating “artificial tipping points through the shrewd selection of structurally influential individuals.” In the experiment at hand, that means getting more moms to adopt best practices for taking care of their babies — and to get them to influence others. Honduras has a high infant mortality rate, and a third of births still take place outside health facilities.
Christakis and Fowler already have some sense of what they expect to find. Results from the team’s preliminary project, conducted in a different region in rural Honduras, were recently published in the Lancet. That study measured villagers’ adoption of new nutritional and water purification programs, and confirmed a strange quirk of social networks known as the friendship paradox. It’s a kind of law of social networks — seen in both offline and online environments — stipulating that when you are asked to name a friend, the friend you nominate has a greater statistical likelihood of being more centrally connected, or having more friends, than you. Highly connected people skew the averages and are more likely to be named.
The challenge so far, however, has been how to quickly find the most connected people to help spread best practices. Presumably, “influentials” in any setting can help reshape these cultural norms, but it is not always obvious who these people are. It’s normally impractical to map out entire social networks to find out.
Christakis and Fowler also want to figure out what percentage of a given community must adopt a practice — 20 percent? 40 percent? 60 percent? — before a new norm becomes universal. “How many people do you have to act on to reach a tipping point?” asks Fowler. “No one has ever asked that, in the field, as far as I know.”
Understanding how to manipulate contagion could empower development work of many kinds, the researchers argue. “Even if you don’t know anything about network theory, you might be able to double your effect,” Fowler adds. “We think these methods travel.”
The Honduras project is driven in part by professional experiences that have left both Fowler and Christakis with something to prove.
In a 2007 study in The New England Journal of Medicine, the pair analyzed Framingham Heart Study data and showed how weight gain could spread through friend networks. Later work also traced how altruism and happiness can spread through friends. Fowler is also one of a relatively small set of outside researchers who have had access to Facebook’s user data, and in 2012 he coauthored a landmark study published in Nature on voting behavior.
Yet the “network revolution” has, on occasion, been dismissed as pop science. It has also failed so far to usher in a lot of social progress. Some in the field argue it is still young — much of the activity has been composed of fundamental research, statistical modeling, early trial balloons, and answering narrow theoretical questions. But its youth has not stopped Silicon Valley from mastering and monetizing the dynamics of peer-to-peer platforms — to great profit. Madison Avenue has found great success in its burgeoning industry of social marketing. And security agencies across the globe have harnessed these techniques.
This gap is because, as Christakis and Fowler note, little of the existing academic social network research has been rooted in the developing world. There are some exceptions. For example, well-known MIT development economists Esther Duflo and Abhijit Banerjee have been studying microfinance networks in Indian villages, while other fieldwork has been conducted, mostly in the domains of economics and anthropology, in places such as Thailand and Peru.
Matthew Jackson of Stanford, who has collaborated with Duflo and Banerjee, says the new work of Christakis and Fowler is “not only of scientific interest but of policy interest,” and he agrees that social science should not limit itself to lab experiments and narrowly construing its mission.
“They are very creative people who seem to always find interesting questions that they can explore — ones that have some application,” Jackson says of Christakis and Fowler. “It’s healthy to say we should be skeptical of things. But we shouldn’t hesitate to approach these problems.”
For their part, Christakis and Fowler acknowledge that the results from the Honduras experiment have limits. The 160 villages they are studying are more isolated than most other areas of the developing world. “I can see people being concerned about generalizability,” Fowler says.
Nonetheless, Christakis stresses that, for many of the world’s poor, the stakes are too great not to innovate and test new methods. “I get annoyed with people who don’t offer alternatives,” he says.
John Wihbey is an assistant professor of journalism at Northeastern University.