Network polarization, filter bubbles, and echo chambers: An annotated review of measures, models, and case studies

Abstract

Polarization arises when the underlying network connecting the members of a society is formed by highly connected groups with weak intergroup connectivity. The increasing polarization, the strengthening of echo chambers, and the isolation caused by information filters in social networks are increasingly attracting the attention of researchers from different areas of knowledge such as computer science, economics, social and political sciences. Despite hundreds of publications in this area, there was little effort to systematize or present the knowledge developed in the field in an organized way. This study presents an annotated review of network polarization measures, models used to handle existing polarization, their applications, and case studies. Altogether, 405 scientific articles and conference papers were examined, with 74 filtered for this review. Several approaches for measuring polarization in graphs and networks were identified, including those based on homophily, modularity, random walks, and balance theory. The models used for reducing polarization included methods that propose edge or node editions (including edge insertions or deletions, and edge weight modifications), changes in social network design, or changes in the recommendation systems embedded in these networks. This review will be helpful to researchers investigating polarized social networks from a theoretical and applied perspective.

Bridging Systems: Open Problems for Countering Destructive Divisiveness across Ranking, Recommenders, and Governance

“…These include metrics based on homophily (the degree to which nodes are connected to nodes that are similar to themselves), modularity (the number of intra-group connections relative to the number of inter-group connections), random walk controversy (a measure of how likely you are to cross between groups when randomly traversing the network), and balance theory (which measures how consistent the network is with properties such as "my friend's friend is my friend" and "my friend's enemy is my enemy"). See Interian et al [41] for a comprehensive review.…”

Voter-like Dynamics with Conflicting Preferences on Modular Networks

“…The result is an unbalanced choice of sources by the agent, who is systematically more influenced by his own community than by the other. In this paper, our main focus resides in determining the level of polarization between the two groups once the system has reached the stationary state, as a function of the preferences' intensities and of the topological structure of the social network [20,21].…”