Social Networks Experiments
Most of us have heard that about six degrees separate us from anybody else on the planet. However, much more important is the observation that our interactions form networks. How interactions are organized is very important – it determines how ideas and diseases spread through society and how our behaviors are shaped. A person infected with the flu is much more likely to spread it to people they interact with daily than to a random person on the street. The network of daily interactions determines whether epidemics will start, and how they will spread.
Just mapping out who knows whom is not sufficient, however. Suppose an acquaintance recommends a movie. Will you go see it? Well, it depends on how much you trust him or her. Moreover, the number of acquaintances that recommend the movie is also important. To know how a movie recommendation will spread through a network of social interactions, it is also important to understand the nature of the interactions. Knowing only the network structure is not sufficient.
This is true in general: both whom you interact with and how you interact determines how ideas spread through our social network. But how do we study the propagation of something as intangible as ideas? Even twenty years ago this would have been a nearly impossible undertaking. Fortunately, online social networks offer an excellent opportunity for scientists to do so.
This approach has been beautifully demonstrated in a pair (here and here with comment here) of studies by Damon Centola from Harvard University. He recruited unknowing participants to the study by designing and advertising a slick looking health website. As people signed up for the site, they were assigned “health buddies”. The experimenter could therefore control who your friends are, and how much they are like you.
Centola made some interesting discoveries by studying how quickly a particular behavior is adopted. He asked participants to join a health discussion forum. Every time they did so their buddies were informed, and asked to join as well. Thus invitations could only travel from buddy to buddy. Propagation was possible only if one of your buddies had joined the forum – there was no other way for you to learn about it.
Interestingly, this behavior seems to propagate most quickly in networks with clustered interactions – where two of your friends are likely to be friends with each other. This may seem counterintuitive. For instance, consider a network where friendships are formed randomly. Now let’s assume we pass a message to a neighbor in this network, and ask her to pass it to all of her neighbors. The message will propagate through such a random network more quickly than through a clustered network with the same number of connections. But Damon Centola discovered that changes in behavior propagate more quickly in clustered networks. How is this possible? The difference is that, as with movie recommendations, we sometimes need to hear a message multiple times in order to adopt a new behavior. And clusters of friends are very effective for such reverberatory propagation.
Scientists are starting to understand how ideas and behaviors spread between us. This could be good news! Hopefully, we will use this knowledge to get people to lead healthier lives, and adopt behaviors that are better for the environment. But these insights can also be used for marketing, and to influence political opinions. We are already being subtly (perhaps not so subtly) manipulated to share information with our circle of friends. By manipulating how and with whom we communicate, advertisers may be able to maximize the spread and the impact of a message. Centola’s work shows that this type of manipulation can be quite effective, and virtually invisible to the participants. I think we should be concerned about such applications.