Topic-Layered Median Consensus and the Emergence of Issue Publics

Logan et al. show the value of induced topic multilayers for Twitter interactions; Berenbrink et al. introduce median aggregation in opinion dynamics. We fuse these to propose “median-seeking link formation”: in each topic layer, agents connect not just to like-minded others, but to influence the local median toward their preference, creating strategic ties that can be heterophilous when a minority can swing the median. This produces issue publics that overlap in surprising ways across topics—coalitions that are transient and cross-cutting even when global homophily is high. We will train GCNN predictors (Ou et al.) to forecast median shifts using layer topology as prior, and use sparsity-based influence inference (Ravazzi et al.) to recover hidden brokers whose cross-topic positioning drives median movement. The novelty is recasting the objective of link formation as median-shaping rather than similarity-seeking, which helps explain unexpected alliances on specific issues despite stable ideological clusters overall. The approach can reveal tactical “issue entrepreneurs” and suggests new interventions: targeting cross-layer median shifters may be more effective for breaking deadlocks than targeting high-degree influencers.

References:

1. Opinion Dynamics with Median Aggregation. Petra Berenbrink, Martin Hoefer, Dominik Kaaser, Marten Maack, Malin Rau, Lisa Wilhelmi (2025). Adaptive Agents and Multi-Agent Systems.
2. Learning hidden influences in large-scale dynamical social networks: A data-driven sparsity-based approach. C. Ravazzi, F. Dabbene, C. Lagoa, A. Proskurnikov (2020). arXiv.org.
3. Social network analysis of Twitter interactions: a directed multilayer network approach. Austin P. Logan, Phillip M. LaCasse, Brian J. Lunday (2023). Social Network Analysis and Mining.
4. Opinion Formation Forecasts in Social Networks: A Graph Convolutional Neural Network Approach. Lizhen Ou, Yiping Yao, Wenjie Tang, Haozhe Yuan, Li-li Chen (2023). IEEE International Symposium on Distributed Simulation and Real-Time Applications.