Researchers' AI study: COVID-19 and Ukraine war synchronized German media

Yara Döring and Felix Bießmann (arXiv:2608.18268) applied large-scale political text analysis to German-language online media, processing several million articles and tweets spanning 2019–2022. Their dataset covered two global crises — the COVID-19 pandemic and the start of the Ukraine war — alongside national political events. Using modern NLP methods for fine-grained political agenda extraction, they compared coverage between German and Swiss outlets, and between newspapers and Twitter. The study, accepted at the SAI4OID workshop (ECML-PKDD 2026), aims to expose algorithm-driven political bias in media recommendations.

The team found that international crises act as a powerful synchronizing force: German and Swiss reporting converged thematically, especially on crisis-driven categories, while domestically influenced topics reflected national policy differences. Newspapers delivered stable, long-form political content, whereas Twitter reacted through short-lived, event-driven spikes. Notably, both mediums still shared a common core during the pandemic. These insights could help media platforms detect bias and inform citizens by making political agendas more transparent, potentially bridging echo chambers and fostering balanced public discourse.