Modeling and Analysis of Correlated, Competing, and Evolving Propagation of Opinions and Information
Abstract
In recent decades, modeling and analysis of spreading processes over complex networks have been considered in a wide range of contexts including cascading failures, epidemics and social contagions, and systemic risk in bank networks, to name a few. Social contagions are typically studied through two different phenomena. Simple contagions, also referred to as information propagation, are used to model cases where a single source of exposure is enough for an individual to get infected and to start spreading the content to their contacts, e.g., news articles, disease spreading, etc. Complex contagions are used to model spreading processes where social reinforcement plays a key role in the spreading process and multiple sources of exposure to a content (e.g., an opinion, a product, etc.) are needed for individuals to change their action or state. Also referred to as influence propagation, complex contagions have been studied in the literature to understand various phenomena including the rise of collective action to join a riot and diffusion of opinions and cultural norms. This project aims to develop new approaches in modeling, analysis, and control of spread of information and influence over social networks. A key observation about the current state-of-the art is that most existing works on modeling influence propagation consider a single content (e.g., an opinion, product, political view, etc.) spreading over a network independently from everything else. However, most real-life examples involve multiple correlated contents spreading simultaneously and exhibiting positive or negative correlation. This project aims to fill this gap in the literature by analyzing the spread of influence over a new contagion model with multiple correlated opinions spreading simultaneously. Leveraging these results, we also aim to investigate the potential impact of correlated opinion propagation on the polarization of the population. We will also study novel influence maximization problems under the new correlated contagion model to develop novel strategies for efficiently controlling the propagation of influence. Other key goals of the project include i) studying non-binary influence propagation over multi-layer networks to better understand the impact of hyper-influencers on shaping publicÕs opinion; ii) developing new methods to predict cases where social media activity can spill-over to physical world and potentially lead to material violence; iii) studying the spread of information over social networks by taking into account the possible mutations or modifications that the information can go through; and iv) developing insights on the spread of misinformation through the information mutation model using data-driven analysis. Successful completion of this research program will require developing new techniques and approaches in the fields of network science, discrete optimization, random graph theory, and percolation theory, together with acquisition and analysis of real-world data. Expected outcomes can help reveal the role of correlations between different opinions on the outcome of influence propagation campaigns and on the level of polarization in the population. They can also help better understand the conditions that lead to a social media activity spilling over to the real-world and can be useful in predicting such events in the future. Finally, our work can help develop new strategies to mitigate spread of misinformation.
Document Details
- Document Type
- DoD Grant Award
- Publication Date
- Aug 02, 2022
- Source ID
- W911NF2210181
Entities
People
- Osman Yagan
Organizations
- Army Contracting Command
- Carnegie Mellon University
- United States Army