Abstract I address the classical problem of opinion homophily in online networks with a case study investigating how topic-level assortativity structures far-right communication on Telegram. Drawing on 61,346 messages from 66 ideologically curated channels, I apply a retrieval-augmented multi-label classification framework and network analysis to examine patterns of topical alignment in message forwarding. Results show that core ideological themes such as ‘Nationalism and white supremacy’ exhibit strong assortativity, forming tightly interconnected clusters, whereas cross-cutting topics such as ‘Immigration’ display low assortativity, linking otherwise distinct groups. Permutation tests confirm the statistical significance of these patterns, revealing that far-right information flows are shaped by selective topical cohesion rather than uniform ideological diffusion.
Yasmine Houri (Sat,) studied this question.