Telegram, the app that some of your friends obnoxiously insist on moving your group chats to – so that you can escape the privacy-evading grip of the “evil “Meta machine – is not exactly an innocent alternative to WhatsApp. While Telegram offers private group chats just like WhatsApp, it also has a large public space that is less hidden, easily accessible, and weakly moderated. This makes it a particularly interesting medium to study.
Thanks to its privacy-enhancing features, Telegram has grown exponentially amid the COVID-19 pandemic and in the wake of stricter content moderation policies put in place by Facebook and Twitter. Hereby, Telegram’s lenient platform guidelines and privacy-focused affordances have been leveraged by extremist groups (e.g., ISIS, Irish ethnonationalists, and far-right extremists) worldwide but also by marginalized communities in authoritarian regimes (e.g., Hong Kong and Russia). As such, Telegram has become more than a tool for political fringe groups and a breeding ground for (online) extremism. Its public space, which I refer to as the Telegramsphere henceforth, forms an important instrument for news broadcasting and news consumption nowadays.
Our study
In our recent study, we were curious (among other things) about the narrative space that characterized Dutch Telegramsphere over time. We studied the Dutch-language Telegramsphere because Telegram usage in The Netherlands received increased media attention and some scholarly attention in the recent past with respect to (1) riots and mass anti-establishment sentiments and behaviour as well as to (2) coordinated disinformation propagation amid the COVID-19 pandemic.
There are two fundamental means to converse on the Telegramsphere: While chats facilitate many to many communication, channels have a broadcasting function (i.e., few to many communication). We collected the full-text message history of 174 channels (N = 94) and chats (N = 80) by querying Telegram’s API via the Telethon Python library from 18 March 2017 until 18 June 2021. Chats/channels were identified via a list of queries relating to a wide range of current affairs in The Netherlands (e.g., news, politics).
To study the development of the narrative space, we conducted topic modelling via BERTopic, a state-of-the-art neural topic modelling method that clusters document snippets into recurring topics. The advantage here over other methods is that researcher input is kept minimal with regards to pre-processing steps, choice of model parameters, and interpretation.
What did we find?
We obtained a total of 125 topics via BERTopic. The figure below illustrates the top 20 topics over time. The most prominent topic was children's safety, which was present throughout the entire investigated timeline and skyrocketed in early 2021. We believe that the climax of this topic may suggest two things: The first, more general explanation is related to an increase in concern with respect to children’s safety during the pandemic. The second, more worrying explanation is that QAnon’s central message ‘save the children' may have permeated the Telegramsphere making users (especially parents) vulnerable to the conspiracy-infused alternate universe, where COVID-19 is a hoax and the real danger that threatens children‘s lives is the deep state. Other prominent topics included vaccination, lies, truth-seeking, religion, elections, police brutality, (fake news) media, protests, war, Donald Trump, and COVID-19.
Taken together the identified topics indicate that the narrative space of the Dutch Telegramsphere has been characterized by anti-establishment and anti-COVID-19 as well as conspiratorial topics, which escalated during the outbreak of the COVID-19 pandemic.
Interested in more insights about the Dutch Telegrampshere or using BERTopic? Feel free to reach out to me (m.simon@uva.nl) for a preprint of my paper that is currently under review.