NSF Award
2418877
Online discourse is shaped by the collective input of users, forming distinct topic clusters with unique interactions and shared interests. Analyzing the development of these topic clusters is crucial for understanding social dynamics. However, most current metrics are designed for static, unweighted, and undirected networks. This project aims to develop interpretable network-based similarity metrics and solvers for dynamic social networks, which can be generalized to multilayer knowledge graphs with directed and weighted edges. By applying these methods to large-scale knowledge graphs from X, covering topics such as COVID-19, QAnon, and hate speech, the research seeks to uncover topic correlations, monitor cluster evolution, and provide insights for societal discourse, awareness, and policy-making. This work supports NSF's mission to promote the progress of science and advance national health, prosperity, and welfare by addressing pressing social issues through innovative computational methods.<br/><br/>The project contributes to two key aspects of dealing with complex social networks. First, it addresses the scarcity of methods that consider directed and weighted edges, which are essential for reflecting the heterogeneous nature of social networks. The project will develop network-based similarity measures based on the Diffusion State Distance metric (DSD), proven effective in weighted and directed networks. This approach draws on similarities between online social networks and protein interaction networks to customize spectral graph metrics for heterogeneous social interaction networks and their multi-layer knowledge graphs. Second, the project aims to enhance the interpretability of community detection algorithms for dynamic social networks. By leveraging the established equivalence between computing DSD and solving graph Laplacian systems, advanced multilevel solvers will be developed using theoretical analysis tools from dynamical systems, differential equations, and spectral theory in numerical linear algebra. This will provide a comprehensive framework for understanding how evolving structures of social networks impact the eigenspaces of the distance matrix. The project includes educational outreach and interdisciplinary collaboration, bridging STEM and social sciences to ensure user-friendly access to the knowledge graph framework and spectral algorithms.<br/><br/>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.
