What a new MIT study suggests about the future of online communities

Sometimes a new scientific study reinforces our previously held beliefs, sometimes it challenges those beliefs and provides new insights. This new study by Damon Centola, an assistant professor of system dynamics and economic sociology at the MIT Sloan School of Management manages to do both.

Centola's study, The Spread of Behavior in an Online Social Network Experiment, appears in this month's issue of Science magazine and has some interesting, and counterintuitive things to tell us about how information spreads through online social networks and how that information may influence actual behavior.Specifically, Centola set up two distinct types of online networks to see which one would be more effective in getting people to sign up for an online health care forum. Using 1,528 subjects he divided them into one group categorized as a 'long-tie' network, meaning an online community whose members are loosely connected with a large number of other members (friends?). In contrast, the second group he set up is called a 'dense cluster', meaning that an individual's connections to others may be fewer in number but denser in the sense that members of this type of group know each other well.

It was expected that information would flow quickly through the long tie network. What wasn't anticipated was the fact that behavior change, the actual act of signing up for the health forum, was slower in the long tie network. People in the dense cluster altered their behavior at a significantly higher rate.
In an article on MIT's site detailing the study they say:

Researchers often regard these dense clusters of connections to be redundant when it comes to spreading information; networks featuring such clusters are considered less efficient than networks with a greater proportion of long ties. But getting people to change ingrained habits, Centola found, requires the extra reinforcement that comes from those redundancies. In other words, people need to hear a new idea multiple times before making a change.

And, finally, they conclude:

People are more likely to acquire new health practices while living in networks with dense clusters of connections — that is, when in close contact with people they already know well.

Centola has made it clear that he considers this study incomplete in some ways and that it really suggests the need to continue with further studies to see how far these conclusions hold up in a more realistic setting. For example, signing up for a health forum may not present a high enough bar to really be representative of how people would behave in real life.

Advertisers, grass roots organizers, church leaders and anyone trying to influence the behavior of large numbers of people have long known that repeated exposure to an idea or product can lead to behavioral change over time. And this study really reinforces that conventional view. In the dense clusters, individuals respond eventually to the repeated exposure to what their 'buddies' were doing. More interesting is the counter intuitive finding that goes along with this: information spreads faster in the long tie networks but actual behavioral change comes slower. That's the key. So, in concrete terms, for anyone designing social media features or deciding how to focus their organization's social media efforts what does this suggest? It may lead us to question the conventional wisdom of the moment--the numbers are on Facebook, yes, but but could it be that our efforts might bear more fruit if they are simply more targeted. Niche social networks may ultimately be more important than the currently huge mega-networks like Facebook. The thousand-friends-facebook-pattern may even be a passing phase.  Real bonds and real shared interests may ultimately count for more than the relative ease and speed of encounters in our presently more popular social media sites. That lasting and meaningful social networks--the dense clusters described in the article- are what keep people together and what allow them to influence each other' behavior is just the perfect case of 'plus ca change'...

See an interview with Damon Centola about the study:

Mainsail through time