We are pleased to share our recent publication in Springer’s Journal of Social Network Analysis and Mining, titled “Modeling Polarized Information Diffusion with SEI(A)I(D)Z: A Stance-Based Epidemiological Approach.” This study addresses a key limitation in prior research on online information diffusion by moving beyond metadata-driven analyses and explicitly modeling the stance expressed in social media content. Recognizing that online discourse is often polarized, the work captures how agreement, disagreement, and skepticism interact simultaneously, offering a more comprehensive view of how competing narratives spread and evolve across digital platforms.
To support this approach, the study introduces a novel stance-based epidemiological framework, SEI(A)I(D)Z, which extends the traditional SEIZ model by differentiating between users who support and those who oppose a narrative. This refinement allows the model to capture interactions among competing viewpoints rather than treating information diffusion as a uniform process. The framework is empirically validated across three distinct contexts and platforms: COVID-19 and 5G conspiracy narratives on X (formerly, Twitter), geopolitical discourse surrounding the Russia–Ukraine war on Telegram, and Taiwanese information campaigns on TikTok during early 2024. Results from non-linear least-squares estimation consistently demonstrate improved predictive accuracy compared to baseline models.
The findings highlight the central role of stance, transmission rate, and user decision-making speed in shaping the reach and intensity of narrative diffusion. In particular, the transmission rate (β) emerges as the most influential driver of the basic reproduction number (R₀), illustrating how increased visibility and virality can accelerate narrative spread. The analysis also shows that slower transitions from exposure to stance adoption extend susceptibility, allowing narratives to persist longer. While manipulative narratives tend to propagate more rapidly, counter-narratives and debunking efforts require stronger systemic and institutional support to achieve comparable impact.
From a practical standpoint, the proposed framework offers actionable insights for policymakers, platform designers, and researchers seeking to curb harmful or misleading content. By explicitly distinguishing between agreement, disagreement, and skepticism, the model enables more targeted interventions, including algorithmic moderation, content throttling, and amplifying credible counter-narratives. Its scalability and adaptability across platforms further position it as a valuable tool for real-time monitoring and informed decision-making in digital governance.
Reflecting on the study, Prof. Nitin Agarwal noted that the “work advances understanding of polarized information ecosystems by explicitly modeling how competing narratives interact and evolve over time. He emphasized that this stance-based epidemiological approach provides a robust and interpretable framework for addressing the complex dynamics of information campaigns, counter-messaging, and skepticism, ultimately supporting more effective strategies to strengthen digital resilience and promote healthier online discourse.”
Click here to read the full article.