Bounded Confidence: How AI Could Exacerbate Social Media’s Homophily Problem

Authors

Dylan Weber, Changing Character of War Centre, Pembroke College, Oxford University and Artis Research
Scott Atran, Changing Character of War Centre, Pembroke College, Oxford University and Artis Research
Rich Davis, Changing Character of War Centre, Pembroke College, Oxford University and Artis Research

Abstract

The advent of the Internet was heralded as a revolutionary development in the democratization of information. It has emerged, however, that online discourse on social media tends to narrow the information landscape of its users. This dynamic is driven by the propensity of the network structure of social media to tend toward homophily; users strongly prefer to interact with content and other users that are similar to them. We review the considerable evidence for the ubiquity of homophily in social media, discuss some possible mechanisms for this phenomenon, and present some observed and hypothesized effects. We also discuss how the homophilic structure of social media makes it uniquely vulnerable to artificial-intelligence-driven, automated influence campaigns.

Recommended Citation

Weber, Dylan; Atran, Scott; and Davis, Rich (2022) "Bounded Confidence: How AI Could Exacerbate Social Media’s Homophily Problem," New England Journal of Public Policy: Vol. 34: Iss. 2, Article 4.
Available at: https://scholarworks.umb.edu/nejpp/vol34/iss2/4