Analysing Anti-Vaccine Concerns on the Gab and Twitter Social Media Platforms

The COVID-19 pandemic intensified longstanding vaccine controversies, highlighting the need to better understand anti-vaccine concerns driving vaccine hesitancy. Most prior social media–based studies classify vaccine discourse into coarse categories (Anti-Vax, Pro-Vax, Neutral), with limited focus on identifying specific hesitancy concerns. Moreover, existing work primarily analyzes mainstream platforms such as Twitter and Facebook, which enforce strict content moderation. In contrast, loosely moderated platforms like Gab—promoting “free speech” and appealing to alt-right users—remain largely unexplored, despite their potential to offer more unfiltered insights. This paper introduces the first annotated dataset of COVID-19 anti-vaccine posts from Gab, where each post is labeled with one or more fine-grained vaccine-concern categories, accompanied by explanations and concern-specific summaries. The dataset supports two tasks: multi-label classification and summarization, evaluated using multiple AI models. CT-BERT achieves the best classification performance (weighted F1 = 1 0.83), while GPT-4 Turbo outperforms others in summarization (ROUGE-S F1 = 0.85, BLEU = 0.82). A comparative analysis of Gab and Twitter further highlights how platform governance influences vaccine misinformation and hesitancy dissemination.