Selective Sharing and the Polarization of Information on Social Networks

Abstract

Online social networks are well known to be politically polarized; regardless of how you map online users – by friends, followers, re-posts or messages – ideologues tend to cluster with co-ideologues. These ideologically polarized networks are likewise believed to be polarized in the information they share – that is, distinct sets of information circulate among different ideological subnetworks. What is not fully understood, however, are the mechanisms that lead to polarized information. Polarization of political information on social networks could, theoretically, be a function purely of “selective exposure;” if users predominantly chose to “follow” or “friend” co-partisans we would expect to likewise see political polarization of information. It is possible, however, that “selectively sharing” – the choices individuals make on what information to forward (re-post, retweet, etc) to their friends – would also influence the degree of polarization, particularly if users share more information that comes from fellow ideologues. This project examines the presence and impact of “selective sharing” in two ways. First, using mathematical and agent based models, I examine the potential magnitude of selective sharing’s effect on information polarization in infinite and complex networks. I also take a preliminary dip into Twitter data to see whether subsequent waves of tweets lead to more or less concentration in one ideological group. The models and brief data foray present contrasting results. Simulating the flow of information in complex networks and likewise computing equilibria distributions of information in infinite networks both indicate that selective sharing results in diminished levels of polarization than would be expected with selective exposure operating on its own. In observing waves of retweets from @FoxNews and @nprnews, however, we see that as tweets get retweeted polarization is either amplified or remains unchanged. The paper ends with a discussion of potential features of social networks – other than selective sharing – that may lead to the observed amplified polarization of information.