Topic homophily in alt-right communications on Telegram

We confront the traditional problem of opinion homophily in online social networks to alt-right communications on Telegram messaging. Using a selected sample of messages from an open-source dataset, we describe the social network of forwarded messages that contain a reference to an alt-tech platform. We build a directed, weighted graph and model the semantic space of their communications to evaluate the topic homophily of the observed groups. We observe a core-periphery network structure with a high imbalance in channel activity. We find that these groups discuss 6 rather distinct topics including religion and identity as well as international politics, media and technologies. We account for topic homophily by computing network assortativity for each topic, and find that assortativity scores are either high when it comes to identity-related conversations, or close to zero when it comes to politics and technologies.