Date of Award
6-4-2026
Document Type
Dissertation
Degree Name
Doctor of Philosophy (PhD)
Department
Information Science
First Advisor
John ohn Talburt
Abstract
Social media has fundamentally transformed how narratives are produced, contested, and disseminated, enabling competing interpretations of the same event to spread rapidly across networked populations. Understanding how narratives gain traction over others in these environments is critical for addressing misinformation, supporting informed public discourse, and designing effective counter-messaging strategies. Yet existing research largely treats narrative propagation as a single-process phenomenon and focuses predominantly on textual data, overlooking both the competitive dynamics between opposing narratives and the growing role of visual symbolic communication in shaping how narratives are received. This dissertation addresses these gaps through a multi-platform, multimodal framework applied to three high-stakes case studies spanning Telegram, X (formerly Twitter), and TikTok. Two research questions guide the study: how visual symbolic elements influence narrative resonance, and what factors drive the propagation of narratives in environments where competing viewpoints coexist. To investigate the role of visual symbols, Social, Cultural and Political (SCP) symbols are extracted from image content using GPT4o and Gemini Pro-Vision, and their effects on user engagement, trust, and emotional response are analyzed. To model narrative competition, this study introduces the SEIAIDZ model, a stance-aware epidemiological framework that extends classical diffusion models by distinguishing between users who agree with a narrative and those who actively oppose it, while incorporating stance-switching, disengagement, and re-engagement dynamics. The findings reveal that symbolic richness is a consistent driver of narrative resonance, with cultural symbols producing the strongest effects on engagement and perceived credibility. The SEIAIDZ model outperforms all baseline models across every platform, demonstrating that explicit modeling of stance competition substantially improves predictive accuracy. Sensitivity analyses identify the transmission rate (β) as the dominant factor governing narrative spread, with the decision rate (ψ) and user exit rate (µ) serving as secondary levers that together determine the basic reproduction number (R0) of a narrative. Counter-narratives are shown to be structurally disadvantaged, requiring sustained effort to achieve comparable reach. These results underscore that narrative dominance in digital environments is shaped not by factual accuracy alone, but by symbolic resonance and structural propagation advantages. The quantified parameters of the SEIAIDZ model provide policymakers and platform designers with actionable, data-driven insights: by targeting β through algorithmic throttling, reducing prolonged exposure via timely fact-checking, and amplifying counter-narrative infrastructure, stakeholders can make evidence-based decisions to mitigate the spread of harmful narratives and foster healthier information ecosystems.
Recommended Citation
Gurung, Mayor Inna, "Modeling the Spread of Competing Narratives in Multimodal Social Media Network" (2026). Theses and Dissertations. 1332.
https://research.ualr.edu/etd/1332
