Drivers of Polarized Discussions on Twitter during Venezuela Political Crisis

Abstract

Social media activity is driven by real-world events (natural disasters, political unrest, etc.) and by processes within the platform itself (viral content, posts by influentials, etc). Understanding how these different factors affect social media conversations in polarized communities has practical implications, from identifying polarizing users to designing content promotion algorithms that alleviate polarization. Based on two datasets that record real-world events (ACLED and GDELT), we investigate how internal and external factors drive related Twitter activity in the highly polarizing context of the Venezuela's political crisis from early 2019. Our findings show that antagonistic communities react differently to different exogenous sources depending on the language they tweet. The engagement of influential users within particular topics seem to match the different levels of polarization observed in the networks.