Experiments with Online Networks

The great majority of traditional work in the social sciences (outside psychology) has involved observational data and not the controlled experiments that are so common in the natural sciences. In 1969, sociologist Morris Zelditch asked, rhetorically, "Can you really study an army in a laboratory?" Nearly half a century later, the answer is "yes." And we do.

In order to build a more unified understanding of the causes of social behavior, we conduct a broad array of experiments, particularly in online settings. We have created a powerful software platform (‘Breadboard’) that allows us to recruit thousands of subjects to come to a virtual lab environment and participate in experiments in which we can manipulate social interactions in diverse ways. Hence, we create temporary artificial "societies" of real people in which we can control and evaluate the workings of social systems. In the past few years, over 20,000 people have participated in our bespoke experiments.

Our general approach is to explore the implications of social network structure and function for collective outcomes, using cooperation, coordination, sharing, and other collective behaviors as model systems. For instance, we have explored how the dynamism inherent to social networks (what we call the "social fluidity" - or the extent to which people can change whom they are connected to), which we experimentally manipulate, may promote cooperative behavior. In other work, we have found that the propensity toward cooperative behavior spreads through social networks, much like other behavioral phenomena we have studied using solely observational data (such as obesity, smoking, or happiness), and we have confirmed mathematical laws governing human cooperation.

Our current agenda is to further examine complex social behavioral patterns that are typical of human sociality, including the emergence of economic inequality; the ability of groups to work together to share resources or to solve problems; and the flow of true and false information in social graphs (for instance, we are interested in the spontaneous emergence of "fake news" and rumors in social systems, and how this may adversely affect groups).