Why a school beats Facebook: how behaviors spread through networks
Article: Why a school beats Facebook: how behaviors spread through networks
This article discusses a study where scientists compare the spread of social behaviors through two general types of networks: one where individuals are connected to others spread throughout the network and another where the individuals are usually connected in clusters. While disease transmission seemed to spread faster within the spread out network, the spread of social behaviors happened more efficiently in the clustered network.
The study used six spread out and six clustered networks. One node in each network was randomly selected to be a member of a health forum, and all of those connected to that individual received an invitation to join. And if one of those individuals joins, then an invite is sent to all of their connections and so on.
It was found that not only did the forum spread faster in a clustered network, it also spread further. In the clustered networks, the behavior spread about four times faster than it did within the spread out networks. And while only 38.26% of nodes from the spread out networks joined, 53.77% of those in the clustered networks wound up joining.
Using the typical notion of the strength of weak ties, one would assume that the spread out networks would propagate the behavior faster since more people would be able to come in contact with the forum, but that wound up not being true.
The researchers suggest that it might be the redundancy of the invites that convinced people to join. If they just got one invite from one friend, they might just brush it off, but if they got multiple invites, they would actually look into it more. And the individuals tend to get more invites in clustered networks where most of their friends are friends with the same people. So the invites would always go out to a similar group of people repeatedly instead of a random group of people each time.
This connects back to what we have heard in lecture about the threshold necessary to adopt a behavior. In the clustered networks, the amount of friends that have adopted a new behavior tends to be higher than in a spread out network because they influence each other. As the proportion of friends that have adopted the behavior increases for one person in a group, it also increases for all the others in the group, driving them all to adopt the new behavior. While in the spread out group, it seems more unlikely for a threshold to be surpassed when they are basically randomly given friends that are not associated with each other.