Polviz: Assessing Opinion Polarization in Social Media through Visual Analytics and Argumentation

GABRIELA DIAZ; DANA URRIBARRI; LUJAN GANUZA; CARLOS IVÁN CHESÑEVAR; ELSA ESTÉVEZ; ANA MAGUITMAN

Pretoria

Conferencia; ICEGOV '24: Proceedings of the 17th International Conference on Theory and Practice of Electronic Governance; 2024

Opinion polarization, a phenomenon characterized by the widening divide between individuals or groups holding contrasting viewpoints, has emerged as a salient feature of contemporary societal discourse, particularly in social media. Opinion polarization poses challenges to decision-making processes, hindering consensus-building efforts vital for addressing pressing different issues such as climate change and socioeconomic inequality. Thus, it is essential to develop new methods and tools for analyzing and addressing the dynamics of opinion polarization. In the last years, visualization techniques have evolved to provide valuable insight when modeling knowledge, and their integration with artificial intelligence techniques has strengthened different possibilities, paving the way for visual analytics and visual knowledge discovery. Computational argumentation is a growing area in AI that has evolved to provide a sound formal model on how arguments are structured, evaluated, and utilized in various contexts, relying on natural language and machine learning techniques. This paper presents Polviz , a novel framework for assessing opinion polarization combining visual analytics and computational argumentation. We provide a characterization of conceptual elements associated with argument-based models (argument, conflict, etc.) oriented towards modeling opinion polarization in social media. We also provide a toolkit of visual analytics tools that allow us to navigate among different models for opinion polarization. Ultimate results corresponding to knowledge emerging from the resulting models are obtained by including a “human in the loop” at the end of the analysis.