Building a Dataset on the 2024 European Elections: How and Why to Study Politics on Short-Video Platforms

While in 2019 Twitter worked as a platform for political communication in the run-up to the European election, in 2024 the growing importance of audiovisual political communication shifted researchers’ focus to short video platforms like TikTok, Instagram, and YouTube. This paper explores the methodological challenges that we faced when building a dataset based on audiovisual data from ten different national contexts on the three short video platforms. We study what circulated and how politics was communicated across the European Union around the 2024 European Elections relying on an innovative methodology that combines automated video scraping with screen recording by 30 researchers in ten countries across Europe. Challenges emerged both in the lack of API access for a lot of these platforms but also the multimodal character of the current formats of communication. Our efforts have not only allowed us to build a database covering all major regions of the European Union (Bulgaria, Croatia, Finland, France, Germany, Hungary, Poland, Portugal, Spain, and Sweden) but also deal with the role of researcher and interpretation in the AI-dominated world. While the video scraping was aimed at gathering data on TikTok, Instagram, and YouTube using account names, hashtags and keywords, the researchers gathered data by recording their mobile phone screens when using organic and synthetic profiles that represented three distinct political ideologies for a month on TikTok and Instagram and writing their reflections on what they could see in the process. The project generated a data set for several research consortia. The videos are categorised in an AI-assisted pipeline for large data use and deep dives. We hope to discuss also the ways in which large datasets can benefit from the interpretive gaze of Large Language Models (LLM). Ultimately, what is at stake in this experiment is the heuristic use of the AI and some theoretical concepts – such as social contract, grievance politics and populism. For comparative political analysis, we seek to demonstrate how “big data” can be approached from a post-foundational and interpretivist perspective, and at the same time yield comparative results – one day with some explanatory power to contemporary political transformations in Europe.