(Dis)connecting people? : examining whether and when social media increases polarization

Abstract

This thesis examines whether and when social media increases polarization. The first empirical chapter looks at the association between social media use and changes in polarization over time. Using data from nine waves of the British Election Study (2015-2019), it shows that self-reported exposure to political information on Facebook and Twitter is associated with increased polarization during each of the last three UK general elections. The second and third empirical chapters explore the conditions under which exposure to political information on social media increases polarization. Using data from two online experiments in the United Kingdom and Ireland, the second chapter shows that exposure to pro-attitudinal information on social media increases polarization, whereas exposure to counterattitudinal information on social media decreases polarization. Furthermore, it finds that the effects of pro- and counter-attitudinal information on polarization are not moderated by the information’s emotional tone. Finally, the third empirical chapter shows that people’s choice of what information to read on social media can have a polarizing effect. Using data from the online experiment in the United Kingdom, it finds that pro-attitudinal information increases polarization to the same extent when people are forced to read the information and when they can choose if they want to read it or not. In contrast, counter-attitudinal information decreases polarization significantly less when people can choose to read the information or not compared to when they are forced to read it. A growing list of political observers worry about the polarizing impact of social media. The thesis contributes to our understanding of social media’s impact on polarization by showing when social media is more likely to increase polarization and when it is less likely to do so.