The study of networks has exploded in recent years. Network scientists have measured, simulated, kicked and prodded almost every social network under the sun.
And in doing so, they’ve discovered all kinds of fascinating properties of these networks, properties that allow them to better understand the spread information and the role that people play in this process. In particular, theorists have long known that better connected individuals play more important roles in a network because they allow information to spread more efficiently. These individuals are of particular interest when it comes to the spread of gossip and even the spread of disease.
It’s easy to imagine that mere humans have always been largely ignorant of these networks, given the complexity of these webs and the complicated measures needed to calculate the connectedness of each node within in them. Which is why theorists have spent significant time and effort in working out ways to find these best connected people. This usually involves mapping the entire network, measuring the connectedness of each individual and then ranking them accordingly, an approach that is time-consuming and expensive, particularly for large networks.
Now Abhijit Banerjee at MIT and a few pals say there is a better way that is faster and cheaper. These guys have discovered that if you ask people to name the individuals in their social network best able to spread gossip, they are remarkably good at identifying them. That’s extraordinary because people have no way of knowing the overall structure of their social network.
Banerjee and co made their discovery by studying the network of links between individuals in 75 rural villages in southwest India. They measured these networks by asking people who they visited, who visited them, who they were related to, who they borrowed money from, who they lent money to, and so on.
They then asked people in 35 villages the following question: “If we want to spread information about a new loan product to everyone in your village, to whom do you suggest we speak?”
The results provide a fascinating insight into the knowledge humans build up about their social networks. When people answered this question (and substantial numbers didn’t), they unerringly identified central individuals within their village.
One possibility is that these people simply named the village chief or the person who lived at the center of the village. But because Banerjee and co already had detailed knowledge of the network, they were able to rule this out.
In fact, the respondents tended to mention people who were more central in the network than either the village chief or the geographically most central person. “This suggests that individuals may use simple protocols to learn valuable things about the complex systems within which they are embedded,” say Banerjee and co.
So how can people know the most central individuals in their network without having any knowledge of the broader network structure? Banerjee and co think they know.
These guys simulated the social network and imagined how an individual within it might work out who was best connected. Network theorists usually approach this problem by adding information to the network at a certain node and then watching how it spreads over time. The node from which it spreads furthest is then the most connected, the best spreader of gossip.
But instead of looking at how information is sent around the network, Banerjee and co simulated how a single node within it might receive information. They then allowed information to percolate through the network and watched what happened.
One crucial factor is that they ensured that the information was like real gossip–that it referred to the specific individual from which it came. For example, “Chad has a new car” or “Fred won the lottery” and so on.
An individual in the network then receives this gossip depending on how well it spreads. Because of this, people can get a good estimate of the connectedness of another individual by simply counting the number of times they hear gossip about them.
The more gossiped about, the more connected that person must be. “This suggests that individuals can rank others according to their centrality in the networks even without knowing the network,” say Banerjee and co.
That’s an important result. Until now, the only way of determining the best connected individuals was by comprehensively measuring the structure of the network and then simulating it to find out how information flowed through it.
The new work suggests a much simpler and cheaper way of getting the same data—simply by asking people. “Eliciting network centrality of others simply by asking individuals may be an inexpensive research and policy tool,” say Banerjee and co.
In some ways, it’s not surprising that humans can come to a detailed understanding of certain properties of their social network without knowing the detailed structure. After all, we’ve lived and evolved in these networks for time immemorial.
Banerjee and co have worked out how and raise the prospect that there might be other simple shortcuts for evaluating networks. Now it’ll be up to the policy makers, the epidemiologist and the marketers to find ways of exploiting this discovery.
So expect to be asked about most gossiped-about person in your network in a survey in the not too distant future.
Ref: arxiv.org/abs/1406.2293: Gossip: Identifying Central Individuals In A Social Network