Funded Grants

The structure and function of social networks

"The earth to be spann'd, connected by network…
The oceans to be cross'd, the distant brought near,
The lands to be welded together"

Walt Whitman, Passage to India (1900)

A century after Walt Whitman wrote Passage to India, the sense of global connectivity that he expressed has metamorphosed from a romantic idea into a deep scientific and social problem. In an era of virtually costless communication, increasingly rapid transport, multinational corporations, international terrorism, and global epidemics, the networks that connect us are too vital, and in some ways too dangerous, merely to be wondered at. This heightened sense of connectivity reflects itself in both the public discourse and academic journals. With several hundred papers published in the last five years, the study of networks is a rapidly developing area of interdisciplinary science, spanning fields from physics to sociology.

Central to this enterprise is a set of as-yet unresolved questions about the structure of large-scale social networks and their role in driving social behavior. How do networks of sexual relationships, for example, differ from networks of friendships or business ties? What psychological and social characteristics influence the formation of relationships in the first place? How important are social networks in shaping labor markets, or enabling business firms to innovate? To what extent is the success or failure of a new technology—or a new book, movie, or TV show—driven by interpersonal versus media influences? In shaping the outcome of events, how important are so-called "opinion leaders" and "connectors" versus otherwise ordinary people who happen to hold opinions?

Social scientists, journalists, and public intellectuals have debated these questions for decades. Yet the answers have proven elusive, for three reasons: empirical data that describes network ties and their influence on behavior is prohibitive to collect manually; the complexity of theoretical models defies both intuition and traditional analytical methods; and large-scale experiments are virtually impossible to conduct in laboratory settings. While these difficulties remain daunting, the power of modern computing and communication technologies is beginning to yield access to the social world as never before. In the era of the Internet, every email sent and every comment posted leaves an electronic trace that, woven together with a multitude of other messages, transactions, and referrals, portrays a vast and ever changing tapestry of social organization from the level of individuals (who are my friends?) to that of whole societies (how is the world connected?). Ten years ago, such a wealth of sociometric data was scarcely imaginable—today it is so commonplace, it passes almost without notice. Yet it presents social scientists with an unprecedented opportunity for empirical research.

By tracking the evolution of the email interaction network of a large university community (comprising about 40,000 individuals), we shall begin to fill in what we see as a significant gap in the theory of social networks. Millions of individual decisions made by thousands of students, faculty, and staff as they enter (or leave) the community, develop (or terminate) associations with one another, and choose which groups to belong to (or avoid), aggregate into macroscopic patterns that are invisible to the individuals who create them and largely independent of personal idiosyncrasies. In studying truly large-scale, evolving networks, we hope to shed light on the interplay between social networks on the one hand, and on the other hand, the social dimensions (race, income, geography) and contexts (academic, extracurricular, residential) that characterize the university experience.

The ability of modern computers to crunch unimaginably large numbers extends beyond empirical analysis—it has also revolutionized our ability to build and explore theories about the world. While it is standard in social sciences to develop models of individual behavior based on intuition, experience, and laboratory experiments, the collective behavior of large groups defies conventional methods. Mobs sometimes riot and loot, but other times remain peaceful. Bad economic news may cause investors to scramble en masse towards the exits, but often leaves them unruffled. Outbreaks of infectious disease sometimes spread globally, but often remain localized. In many cases, the difference between wildly divergent outcomes derives not from the characteristics of the individuals themselves, but from their interactions. A critical component of our project is the development of mathematical network models. These models, while still quite simple to write down, exhibit features that are beyond the predictive capabilities of conventional pencil-and-paper mathematics; hence we simulate their development on computers, measuring their statistical properties and comparing them to realworld data.

Finally, the technology of the Internet affords us the capability to conduct experiments on a scale far larger than is possible in the traditional laboratory setting. Inspired by the early work of social psychologist Stanley Milgram, we have adapted his "small world method" to the Internet ( By studying what we call "social search"—in which individuals search for remote "targets" by forwarding messages to acquaintances whom they consider closer to the target—we can explore the role of social networks in facilitating purposeful action. What is remarkable about social search is that it is fundamentally a collective activity: no individual has enough information to direct a message all the way from its origin to the target, but chains succeed anyway; hence the chains somehow "know" more than the individuals who create them. While the intelligence required to conduct social search still resides in people, the key to harnessing it lies in the network. By conducting a series of social search experiments, we hope to understand better the structure of global social networks, and also how individuals perceive and manipulate that structure.

Networks are a critical part of social reality—carrying information, transmitting disease, and providing access to resources—and they have long been recognized as such. But if we are to penetrate the surface of this observation, it will require a coordinated approach between theory, simulation, data, and experiment. We propose to take just such an approach, and in so doing contribute not only to network analysis, but also more broadly to social science and policy.