Using social network analysis to study asymmetric polarisation on Facebook

This study focuses on asymmetric polarisation, a concept used to describe polarised contexts where one of the sides of the political divide is more extreme than the other and often engages in problematic behaviours, such as relying on partisan media and sharing mis- and disinformation. To examine asymmetric polarisation, this research relies on a dataset of 481,610 Facebook posts containing URLs from the 2022 Brazilian election run-off. Social Network Analysis metrics of centrality (in- and out-degree) were used to identify the most active Facebook entities and the most shared URLs in the dataset, and a community-discovery algorithm (modularity) was used to identify clusters of entities that more often shared the same or similar URLs. A manual review and classification of the most active entities and most shared URLs was used to make sense of the network and explore the potential asymmetric dynamics of the link-sharing network on Facebook. Findings show that the most shared URLs within the pro-Bolsonaro cluster include links to misleading content from partisan outlets and to a website created for Bolsonaro supporters to oversight the voting to avoid allegedly election frauds. There are also links to Instagram posts from Bolsonaro and his political allies sharing highly partisan and misleading content. On the other hand, most of the prevalent URLs within the proLula cluster linked to mainstream media, although some linked to campaign websites and social media posts. This shows evidence of an asymmetric polarisation dynamic during the election.