Identity-Based Attribute Prototypes Distinguish Communities on Twitter

Abstract

This paper examines the link between conversational communities on Twitter and their members' expressions of social identity. It specifically tests the presence of community prototypes, or collections of attributes which define a group through meta-contrast: high in-group cohesiveness and high out-group distinctiveness. Analyzing four datasets of political discussions ranging from roughly 4 to 30 million tweets, we find strong evidence for the presence of distinctive community prototypes. We observe that community prototypes are constructed through hashtags, mentions, emojis, and identity-phrases. This finding situates prior work on the identity signaling of individual users within a larger group process playing out within communication communities. Community prototypes are then constructed for specific communities by measuring the salience of identity signals for each community. Observed community prototypes tend to be based on political ideology, location and language, or general interests. While the presence of community prototypes may be a natural group behavior, the high levels of contrast observed between communities displaying ideologically opposed prototypes indicate the presence of identity-related polarization.