Polarization in Social Media Discussions during Elections and Referenda

• In Event: Campaigns 2 - Advertising and Online Politics

Abstract

The spreading of online propaganda (including misinformation and disinformation) by human “trolls” and bots has recently received public attention, particularly because of its possible effect on the outcomes of elections and referenda. As this effect cannot be directly observed, we have to resort to proxy measures to derive causal inferences. One of these measures is the polarization of discussion in social media, as we would expect an increase in polarization driven by trolls and bots. In this work, we undertake a comparative study of trolls and bots and their influence on polarization across various elections and referenda on Twitter. We examine both qualitative and quantitative differences between the online discussion across several case studies of elections and referenda in 2018, including the Irish abortion referendum, Macedonian (FYROM) referendum, Brazilian presidential elections and the US midterms. Specifically we study the re-tweeted and co-retweeted network of Twitter users participating in contentious political discussion. The scope of this study includes activity of Twitter accounts potentially tied to foreign states. We quantify the influence of bots by using graph-theoretic properties of connectivity and force-directed group distance. One of the advantages of this work is that its methods can be easily reproduced to study forthcoming elections and referenda across different nations and linguistic communities, as well as contribute to comparative studies of how automated propaganda influence public opinion in non-election contexts.

Authors

• Abhishek Dasgupta, University of Oxford

• Monica Kaminska

• Panagiotis Metaxas, Wellesley College

• Lucas Kello, Oxford University