Lessons learned from collecting and managing an online communication dataset from right-wing extremist actors

Data from large online communication platforms enable researchers to study a variety of specific communication settings. But archiving and sharing data used for this kind of research is a challenge on its own. And it becomes even more challenging for particularly sensitive data. We present a use case for creating a dataset from online communication of German and Austrian right wing extremist actors.

While many studies focus on a specific platform, powerful actors in the right-wing scene are connected by an online ecosystem encompassing multiple platforms. Often, hateful content is linked to on other platforms and actors tailor their language according to platform affordances (e.g., the level of content moderation). We have created a cross-platform dataset with data from Telegram and YouTube, that is based on a curated list of Telegram channels by Austrian and German right-wing extremists and that also includes the outgoing links from these actors’ Telegram posts to the mainstream media platform YouTube.

We will describe details of our procedures for planning, collecting, managing, annotating, and archiving the dataset. This includes lessons learned from seeking ethics approval. It also includes reflections on strategies for collecting data during a time of restricted public access to online platform data, and for managing the data during the collection process. It furthermore includes reflections on protecting researchers involved in the data collection, including student assistants who are helping with data annotation to prepare data for analysis. In order to be able to archive and share the dataset, we had to face the trade-off between ethical and legal concerns and data quality, resulting in several limitations of the final dataset. We summarize aspects that we consider useful as guidance for other work in this areas and that can be transferred to other cases of sensitive content from online platforms.