What they do in the shadows: analyzing the networks of far-right actors on Telegram
Aleksandra Urman, Stefan KatzThe present study examines the interconnections between Western far-right actors on Telegram messenger. Far-right actors rely on online platforms to promote their messages and bypass the gate-keeping of traditional media. Twitter and Facebook had long been the platforms of choice for these actors who, after being banned by there, were forced to “migrate” to other platforms. According to media reports, in 2019 far-right flooded the more privacy-focused Telegram. Though Telegram is primarily a messenger, one can also create “Public channels” there. Messages posted to such channels can be seen by all Telegram users, forwarded to other chats and channels, but not commented on.
We analyze the connections between the public channels of Western far-right activists and groups. The data was collected through the Telegram’s API via the Telethon Python library using snowball sampling. We collected the messages of far-right public channels from September 2015 to November 2019 and the channels cited by them, and then constructed a citation network based on that. The full directed network (collapsed into one snapshot including all the citations over the aforementioned period) consists of 39109 nodes and 144511 weighted edges (the weight of an edge is equivalent to the number of citations).
The paper aims to answer two research questions. First, we examine whether the community structure of the network is similar to the structures of the far-right networks previously examined by scholars on Facebook and Twitter. Following the methodology of the studies on other platforms, we apply the Louvain community detection algorithm to the final snapshot of the network. We find that the Telegram far-right network is less centralized than the networks on Facebook and Twitter. The far-right core of the network is mostly fragmented along two lines: linguistic (with distinct English-speaking, German-speaking and Italian-speaking communities) and ideological (with the distinct English-speaking alt-right, libertarian and national-socialist communities). In addition to the far-right core, there is a part of the network dedicated to memes and counter-culture. Second, we try to identify the tipping points in the dynamics of network formation. For this, we first examine the ratios of edge formation in the network by month. We find four major spikes in the rate of the growth of the network, each coinciding in time with the bans of far-right actors on Twitter and Facebook. We also aim to establish which non-far-right channels act as brokers for the entry of the far-right actors to Telegram. For that, we divide the network into 4-month snapshots and then examine community structures and descriptive statistics (e.g., node centralities) for each of them. We find that Telegram channels related to 4chan and 8chan image-boards act as major brokers between the far-right and meme and counter-culture communities. These brokers helped far-right actors greatly amplify their messages at the early stages of the far-right network development.