Characterizing network dynamics of online hate communities around the COVID-19 pandemic

摘要: Hate speech has long posed a serious problem for the integrity of digital platforms. Although significant progress been made in identifying hate its various forms, prevailing computational approaches have tended to consider it isolation from community-based contexts which spreads. In this paper, we propose dynamic network framework characterize communities, focusing on Twitter conversations related COVID-19 United States and Philippines. While average scores remain fairly consistent over time, communities grow increasingly organized March, then slowly disperse succeeding months. This pattern is robust fluctuations number clusters cluster size. Infodemiological analysis demonstrates that both countries, spread around features similar reproduction rates as other information Twitter, with spikes generation at time points highest community-level organization speech. Identity further reveals US initially targets political figures, grows predominantly racially charged; Philippines, consistently time. Finally, demonstrate higher levels community are associated smaller, more isolated, highly hierarchical across contexts. suggests potentially shared structural conditions effective online even when functionally targeting distinct identity groups. Our findings bear theoretical methodological implications scientific study understanding pandemic’s broader societal impacts offline.