It’s thought that every person in the United States is connected by no more than six degrees of separation.

Proving it, though, is another matter.

For the most part, computer scientists, mathematicians, sociologists and other researchers have worked independently to map the vast economic, political and social networks that help us decide when to buy or sell a car, who to vote for, or even which foods we like and dislike. But an emerging field of research called “network science” is bringing all of these disciplines togther.

The goal is to better understand, predict and design the behavior of the networks that link and shape us.

Penn’s Michael Kearns is at the forefront of the movement. A professor in the Department of Computer and Information Science, his most recent study has shown how networks can help generate the global adoption of minority viewpoints.

“What’s driving this is more and more network interaction happening in a way that can be measured,” Kearns says.

Computer scientists and sociologists, for example, can benefit from studying social networking Web sites like Facebook. E-mail and instant messaging data can be mined to map who is talking to whom. And technological advances made by neurologists collaborating with mathematicians are helping map the vast network of neural activity in the brain, leading to promising new risk assessments and disease therapies.

“These radically different networks all have undergone a measurement transformation,” Kearns says. “We now have hard data for the first time.”
In his recent study, Kearns’ group seated 36 test subjects in a room at computer terminals. Each computer was linked to between two and 18 others in the room. None had a global view of the overall network, and no communication was allowed.

Users were given a set time to vote for either red or blue. If everyone in the group could agree on the same color within one minute, everyone was rewarded play money. If they failed to reach consensus, they got nothing.

However, some participants were rewarded greater amounts of money depending on the color that won—red could mean a bigger payout to some, blue to others. This created tensions between private incentives, global unity and the structure of the network.

Despite conflicting motivations, however, consensus was reached and subjects rewarded in 55 of 81 experiments.

Kearns’ study showed that the structure of the network alone could influence the outcome. Put simply, even a super-minority could rule, provided it gains enough influence (or connections) to shape popular opinion. In one experiment, as few as six subjects who preferred red were able to convince 30 others preferring blue to vote with them and reach a consensus.

Other common influential factors like rhetoric, emotion and advertising played no role in getting people to change their vote. This scenario played out in the 2008 Democratic primary race, where fears of a split vote ultimately led some supporters of Hillary Clinton to back Barack Obama instead.

“I’m trying to boil it down and show that this might be purely an anonymous interaction within a network,” Kearns says.

The study was performed by Kearns in conjunction with Stephen Judd, Jinsong Tan and Jennifer Wortman of Penn’s Department of Computer and Information Science.

Future studies will move from the lab to the Web, where scientists will continue to study the relationship between network connections and the way we act, both individually and collectively.

“The prospect of semi-controlled experiments on a massive scale is around the corner,” Kearns says. “We’re starting to see fragments of it already.”

With more than 200 million voluntary users worldwide, Facebook seems an obvious jumping off point. “I think we’ll soon see an explosion of Web-based social network experimentation built on top of existing or new platforms,” Kearns says.