Many civil society and academic projects are engaged in documenting and analyzing conspiracy theory, anti-democratic or racist discourse on social media. Kilian Buehling has investigated to what extent the results of these efforts are influenced by the time of data collection. 

When dealing with any social media data, researchers are advised to always consider the extent to which message deletions may systematically bias their own results. In the recently published study “Message Deletion on Telegram: Affected Data Types and Implications for Computational Analysis,” a sample of right-wing conspiracy theory Telegram chats was examined for subsequent deletions. For this purpose, 25 channels and 25 groups were queried every half hour to register all new posts and to anticipate possible deletions. We aim to assess the extent to which deletions can lead to systematic biases in research results by making this survey as complete as possible.

A first overview showed that, on average, only 88% of the sent messages are available after five days. After seven months, only 83% are still available. In chat groups, even more messages are deleted: After five days, only 64% of the messages are still readable, and after seven months, it is only 52%. This discrepancy may caused by using moderators and chat bots in the groups. The types of messages that are deleted also differ between these two types of public communication: In chat groups, it is primarily text messages that are deleted. In channels, a disproportionate number of shared messages from other channels and videos are removed.

For further information, you can read the full paper, which is available as an open-access publication. The study was developed within the NEOVEX project, which was funded by the German Federal Ministry of Education and Research (BMBF) as part of its “Research for Civil Security” program.