Social media has become a central battleground for political and ideological debates, influencing real-world conflicts and crises. However, traditional methods for measuring online polarization and radicalization struggle with scalability and data limitations. This thesis leverages recent advancements in Natural Language Processing (NLP), particularly transformer-based models and Large Language Models (LLMs), to develop more efficient and accurate tools for analyzing polarization. It introduces novel techniques to quantify echo chambers, examine cross-partisan interactions, and assess radicalization in gender-based communities. Additionally, it enhances stance detection by fine-tuning language models to differentiate opposing viewpoints more effectively. These contributions provide Computational Social Scientists with powerful new methods to study and mitigate online polarization.

About Vahid Ghafouri

Vahid is a Ph.D. student in Telematics at IMDEA Networks Institute (+UC3M). His main area of research involves the application of NLP on social network data for measuring online polarization and radicalization.

Supervisor de tesis: Guillermo Suarez-Tangil, IMDEA Networks Institute, Madrid

Universidad: Universidad Carlos III de Madrid, España

Programa de doctorado: Ingeniería Telemática

Miembros del tribunal:

• Presidente: Dong Nguyen
• Secretario: Rubén Cuevas, Associate Professor in the Telematics Engineering Department at Universidad Carlos III of Madrid (UC3M)
• Miembros del panel: Walid Magdy, Associate professor at the University of Edinburgh