Hong Kong and Catalonian protest mobilization networks on Telegram: evolution and change
Stefan Katz, Aleksandra UrmanIn this paper, we analyze two protest movements of 2019: the Catalan protests and the Hong Kong protests. Both movements rely heavily on the Telegram messaging app for spreading their messages, coordination and mobilization. While Telegram is primarily a messaging app, it also allows users to create public channels to use as newsfeeds. Information published in these channels can be accessed by all Telegram users, forwarded to others but not commented on. Hence, Telegram can be called a forwarding network.
We analyze the structures of such forwarding (or citation) networks between Telegram public channels and public group chats related to the two protest movements. We collected the data via Telegram’s API using snowball sampling starting with two seeds – the most popular (by the number of subscribers) public channels related to Hong Kong protests (@Dadfindboy channel, 239 thousand subscribers) and Catalan protests (@tsunamid, 406 thousand subscribers). We collected all the messages posted by these channels, then extracted all the mentions and citations of other channels from them, collected the most frequently cited channels and repeated the procedure for thousands of other channels. Thus we were able to gather data even on small channels and in this respect, the network is quite complete. However, since the protests in both cases are ongoing, we keep updating our library of collected messages to include new posts as well.
In this research, we aim to answer three main research questions. First, we examine the topologies of the citation networks for both movements (including messages prior to the start of the protests – June 2019 for Hong Kong and October 2019 for Catalonia) to find similarities and differences between the two. We look at overall centralization, node centralities and distribution of communities as well the influence of foreign groups within the networks. We find that the network related to Hong Kong is much more centralized than the Catalan one. The Catalan network is more closely connected to foreign actors, including several Western far-right groups, while the one in Hong Kong has few connections to foreign groups and channels. Second, we identify the most prominent channels in each movement (e.g., by looking at node centralities and hubs and authorities in the network). We find, for instance, that the Catalan wing of the Anonymous hacker group has been one of the most influential actors in the Catalan protest mobilization network. Finally, we aim to establish the dynamics of network formation and change for the two networks. For that we include temporal dimension and a) compare the network structures (including community structures) before and after the start of the protests; b) divide the post-start of the protest networks into several snapshots (each corresponding to a period of two weeks), perform community detection and run descriptive statistics analysis on each of them. Comparing network dynamics allows us to identify potential shifting of influence between subgroups in the network which in turn can be indications for shifting objectives and approaches within the protest movements.