How Facebook 'Contagion' Spreads
To join Facebook or not to join Facebook? You might think the decision depends on how many of your friends are already on the social-networking site. But a new study reveals that it's not the raw number of friends that matters but rather the types of friends who are signed up. The results are the first to show that groups of friends—rather than friend number—are important to how social trends spread.
Previous research on how people make decisions—whether to buy a product, attend a show, or pick up a new hobby—has concluded that the more people you know who do something, the more likely you are to do it. But that's not what computer scientist Jon Kleinberg of Cornell University and colleagues found when they started analyzing data on decisions to join social media networks.
The team collected data from 54 million Facebook invitation e-mails. When people sign up to use Facebook, they are given the opportunity to send invitations to anyone in their e-mail address book who is not already a member of the site. The invitation includes the name of the member who initiated the e-mail as well as any the names of Facebook user who has previously imported the recipient's e-mail address. So with each invitation someone gets, their list of potential friends on subsequent messages grows. The data the researchers—who included a collaborator within Facebook—analyzed included the friends listed, those friends' demographics and connections within the Facebook network, and whether the invitee joined the site. The team also used data on how frequently new Facebook members ended up using the site over the next 3 months.
"What jumped out at us was that someone's likelihood of joining really corresponded not to the number of friends represented, but to how many disconnected groups the friends listed on the e-mail fell into," Kleinberg says. "We were surprised by how clean the effect was."
If four people who were all connected via Facebook friendships were listed on the invitation, for example, the recipient was as likely to join the site as if one friend was listed. But if the message contained the names of four people who had no direct Facebook friendships between them, the odds of the recipient joining the site more than doubled. Moreover, new Facebook members whose first 20 friends fell into multiple distinct groups were much more likely to stay engaged with the site—visiting Facebook 6 of 7 days a week 3 months after registering—than those who had 20 friends all within one connected group, Kleinberg's team reports online today in the Proceedings of the National Academy of Sciences.
"Researchers spend all this time studying what makes someone be an influencer, and this paper is great because it really turns that on its head to see what causes people to be influenced," says James Fowler, a social scientist at the University of California, San Diego, who was not involved in the study.
As for why people are more likely to join Facebook if multiple groups of friends, rather than simply a large number of friends, already use the service, Fowler has a couple of ideas. Maybe, he says, one group is the most important in someone's decision on whether to join the site. In this case, listing friends from multiple groups increases the odds that this key group will be represented. "Maybe you don't care at all about whether your friends from high school are on Facebook, but when you see someone from your jujitsu club, that makes the difference."
Or, Fowler says, it could be that having more groups of friends using Facebook increases the pressure on a person to join, the same way having many friends doing something has been shown to have an effect on other behaviors. Either way, he says, future studies on how friends on social networks influence behavior will likely delve more into the effect of different friend groups.
"The biggest question remaining is whether this will translate into real-world behaviors," Fowler says. If multiple Facebook friends of a person show an interest in a product, for example, is that person more likely to buy the product if those friends belong to different social groups? And what about the world beyond Facebook? Fowler cautions against extrapolating the results: "I think that we're going to start seeing that the way online behavior is influenced is not always the same as how real-world behavior is influenced."