Measuring the epistemic asymmetry of polarized Twitter comunities through network analysis. The case of COVID-19 in the US, Australia and Argentina.

Political polarisation over science is the phenomenon by which individuals form beliefs about science- related topics following partisan lines rather than scientific consensus and available evidence. The asymmetry thesis holds that right-leaning political ideology correlates with a set of psychological and epistemic features that makes agents worse at information processing tasks than their left-leaning counterparts. Detractors of the asymmetry thesis argue than when a topic becomes polarised, cognitive biases, such as politically motivated reasoning, are triggered in equal measure across the political spectrum. In this paper we take on the discussion looking for the asymmetry where most individuals seek, evaluate and transmit information and where most polarised discussions live: on social media. Taking as a case study the discourse on COVID-19 in Australia, Argentina and the US we conduct a 3-country study spanning 11 months of Twitter interaction with over 108 million tweets. We shift the focus from individuals to communities using social network analysis using the wisdom_of_crowds library which we re-implemented in C++ to scale the implementation to networks of more than 600,000 nodes and 10 million edges. We found that the left-right asymmetry holds across the 3 countries in the sample despite the differences in language, party system structure, and polarisation levels. Remarkably, we found that, while there is a significant epistemic asymmetry in left vs right communities, it is concentrated in a small subset of each group. This offers a possible explanation to the current stalemate in the asymmetry thesis debate, since, while left-leaning groups had overall better epistemic position, the majority in all communities showed a rather poor epistemic behaviour