Automated election audits - Analyzing Exposure to Political Content on Social Media with a Case Study of TikTok in Germany’s 2024 Regional Elections
Year: 2024
Type: preprint
Abstract: A main obstacle to the study of social media's impact on democracy is limited access to platform data. We present sock-puppet audits as a method for political scientists to collect social media data building on the algorithm audit literature in computer science and experimental approaches in the social sciences. Data is collected by scraping feeds of automated users (bots) of varied profile and behavioral characteristics. Our framework produces real-time data suitable for both quantitative and qualitative analysis without relying on platform-controlled APIs. First, we demonstrate the step-by-step execution of sock-puppet audits using pseudocode. Second, we apply the approach to a case study on political content exposure on TikTok during Germany’s 2024 regional elections (N=230.000 videos). Within just three weeks on the platform, we find that new users without any political interest have a 3-4 times higher likelihood of being exposed to the right-wing populist party (AFD) relative to mainstream parties. We further discuss cross-platform opportunities for application of this approach as well as legal, ethical and technical challenges.
Authors: Jasper Tjaden, Johannes Wolfgram, Aaron Philipp, Sarah Weissmann, Licia Bobzien, Ulrich Kohler, Roland Verwiebe
Institution: University of Potsdam
Language: English
Topic: Social Media and Politics
Subfield: Communication
Field: Social Sciences
Domain: Social Sciences
Open Access status: green
Funder: Universität Potsdam
OpenAlex ID: W4404940754