Existing studies of political polarization are often limited to a single country and one form of polarization, hindering a comprehensive understanding of the phenomenon. Here we investigate patterns of polarization online across nine countries (Canada, France, Germany, Italy, Poland, Spain, Turkey, UK, USA), focusing on the structure of political interaction networks, the use of toxic language targeting out-groups, and how these factors relate to user engagement. First, we show that political interaction networks are structurally polarized on Twitter (currently X). Second, we reveal that out-group interactions, defined by the network, are more toxic than in-group interactions, indicative of affective polarization. Third, we show that out-group interactions receive lower engagement than in-group interactions. Finally, we identify a common ally-enemy structure in political interactions, show that political mentions are more toxic than apolitical mentions, and highlight that interactions between politically engaged accounts are limited and rarely reciprocated. These results hold across countries and represent a step towards a stronger cross-country understanding of polarization. Identifying patterns of polarization is important for understanding its root cause. Here, using Twitter data from 9 countries, the authors show that out-group mentions use more toxic language than than in-group mentions, and political mentions are more toxic than apolitical mentions.
Stored only when both papers are in the corpus — 35 citations exist upstream.
Imported from OpenAlex (CC0) — nothing here is generated.