Penn researchers' Social Influence Game models adversarial persuasion 10x faster

A team of researchers from the University of Pennsylvania has introduced the Social Influence Game (SIG), a new framework designed to model adversarial persuasion in social networks. The SIG allows an arbitrary number of competing players to allocate influence from fixed budgets, steering opinions that evolve under DeGroot dynamics. The team proved the resulting optimization is a difference-of-convex program, enabling tractable analysis. To achieve scalability, they developed an Iterated Linear (IL) solver that approximates player objectives using linear programs. In experiments on random and archetypical networks, the IL solver achieved solutions within 7% of nonlinear solvers while being over 10x faster, scaling to large social networks. This work lays a foundation for asymptotic analysis of contested influence in complex networks and was accepted at the American Control Conference 2026.

The SIG framework captures structural leverage points of networks, allowing players to strategically choose which nodes to influence. By modeling contested influence—where multiple actors vie for opinion dominance—the paper addresses a real-world challenge in social media, political campaigns, and misinformation spread. The IL solver's speed and accuracy make it practical for large-scale simulations, potentially aiding researchers and policymakers in understanding how influence campaigns unfold. The authors note that future work could extend the model to dynamic budgets, adaptive strategies, or heterogeneous influence mechanisms. This research bridges control theory, network science, and multiagent systems, offering a rigorous yet interpretable tool for analyzing persuasion in contested social networks.