UMass Machine Learning and Friends Lunch | Main / Social Influence Over Social Networks

This talk will present two very different approaches to the study of social influence over social networks. In the first case, there is informational influence spreading through an artificial network. In the second, there is normative influence over a real friendship network.

In the first part of the talk, I will present results from an experiment with human subjects in which networked groups are searching a problem space and sharing information about their search. Participants had 15 rounds to guess a number between 0-100, and received a score based on a continuous fitness function. In the next round they saw their score as well as the guess and score of their network neighbors. We varied the network structure and the fitness function, and observed how quickly the groups converged on the optimal value.

In the second part of the talk, I will discuss the results of a study examining the extent to which political attitudes are shared among friends. We created an application for Facebook that asked users to answer questions about their political attitudes and to guess how their friends would answer the same questions. We looked at the degree of true homophily (similarity among friends), perceived homophily, and the accuracy of friends' perceptions of each other.