Abstract
The proliferation of digital social networks has intensified ideological polarization and created new arenas for information warfare. We develop a mathematical framework for strategic influence competition between two (or more) opposing groups in a network of opinion setters, each represented by a disjoint set of influencers. For a randomly selected social media user, we model the evolving identity of the most influential opinion setter using a Markov chain. From the perspective of one group, we maximize the discounted cumulative probability that this opinion setter belongs to that group, capturing the tradeoff between immediate and long-term influence. The transition matrix serves as the decision variable and is constrained by finite manipulation budgets. Although the resulting problem appears intractable at first glance, we derive a convex, and in some cases linear, reformulation through suitable variable transformations. Experiments on real-world U.S. political datasets show that the proposed model can reverse the baseline influence outcome, enabling a group with initially weaker influence to achieve dominance.

This publication has 14 references indexed in Scilit: