Who Shapes Brazil's Vaccine Debate? Semi-Supervised Modeling of Stance and Polarization in YouTube's Media Ecosystem

Researchers from Brazilian universities (Geovana S. de Oliveira, Ana P. C. Silva, Fabricio Murai, Carlos H. G. Ferreira) developed a semi-supervised stance detection framework combining self-labeling and self-training to classify nearly 1.4 million YouTube comments on Brazil's vaccine discourse. Published on arXiv (2604.18586) and accepted at WebSci'26, this is the largest longitudinal study of vaccine debate in a non-English, high-impact context. The model analyzes comments across Brazil's full immunization schedule, integrating temporal patterns, engagement metrics, and channel taxonomy including legacy media, science communicators, and digital-native outlets.

The results show that semi-supervised learning significantly improves stance classification robustness, enabling fine-grained tracking of public attitudes. Polarization spikes during epidemiological crises, particularly COVID-19, but becomes fragmented across different vaccines and interaction patterns in the post-pandemic period. Notably, science communication and digital-native channels emerge as primary loci of both supportive and oppositional engagement, revealing structural vulnerabilities in contemporary health communication. The study provides actionable evidence for public health agencies, platform governance, and online information ecosystems.