Opinion Formation and Graph Dynamics: From Modeling to Empirical Applications

Arizona State University, Scottsdale AZ

Investigators

Abstract

The rise of social media has drastically changed our conception of opinion formation. Rather than a few central outlets (TV channels, newspapers) acting as a common and external frame of reference, many now derive their information through interaction on social platforms. Therefore, the dynamics of opinion formation have become more self-organized where large scale behaviors emerge as a result of local interactions without a central authority, similar to a flock of birds or a shoal of fish. But how are opinions formed in social networks? Is the structure of social networks a reaction of the users' opinions or does the network shape the users' opinions? One of the key motivations for this project is the following paradox: while the use of social media has supposedly increased the connectivity among individuals, opinions have become more polarized; homophilic "echo chambers" appear to be a ubiquitous feature of the structure of social media. These observed behaviors call for a close examination of the interplay between social media interaction and the dynamics of opinion formation at scale in order to better understand the rise of polarization and echo chambers. This project will also provide research and mentorship opportunities for graduate students. As an emergent behavior, the underlying dynamics of opinion formation is unknown and cannot be derived from physical principles. Thus, the framework developed in this project is interdisciplinary at its core and combines both mathematical modeling (dynamical systems, graph theory) and data acquisition and analysis (natural language processing, sentiment analysis, clustering algorithms). This project will investigate novel dynamical systems modeling the interplay between a graph and a system of opinions. The aim is to characterize conditions resulting in polarization, consensus, or a transition between these two states. Then, experimental data will be extracted from the major social media platforms using a variety of data capture techniques. The resulting measurements will allow for an investigation of how the network structure of social media and the distribution of its user's opinions change overtime, providing insight into the interplay between these two dynamics. The ultimate goal is to refine the modeling pursuit with experimental data in order to investigate what is driving the observed trend towards polarization. Finally, using models validated with empirical data, the project will investigate various strategies aimed at reducing polarization and enhancing the natural flow of various opinions within a large audience. This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.