Measuring Polarization, Moderation, and Destigmatization on Social Media

• In Event: The Perils and Promise of Social Media Discourse and Information

Abstract

Political scientists still have limited tools to extract political meaning from short texts, which poses challenges to the study of polarization in online communities. As an alternative to text analysis, I propose a repeated cross-sectional approach that combines two stages of network analysis to detect changes in mass discourse on social media. This cross-sectional approach features before-after comparisons made more valid with covariate conditioning (on network ideology). Once cross-sectional comparisons are constructed in the first stage of network analysis, I leverage the network ideology of who is being retweeted by whom to proxy for in-network discursive shifts. I describe how this two-stage network analysis makes it possible to capture three discursive shifts: (1) polarization, (2) moderation, and (3) destigmatization. Then, I use it to examine the effects of high salience news events on Twitter discourse. I find that in day-to-day social media discourse, Twitter users across the ideological spectrum express predictably partisan opinions on a range of political topics. Following major news events salient to those topics, sampled users' retweet networks sometimes moderate, sometimes polarize, while at other points, users across a broad range of ideological communities retweet others situated in more extreme networks. I refer to this last type of shift as networked destigmatization. I conclude by discussing how this two-stage framework can help political scientists understand the effects of politicized events on public attitudes.

Author

• Laura Jakli, Harvard University