Evolution of discordant edges in the voter model on random sparse digraphs

TL;DR

This paper investigates how opinions evolve over time on directed random graphs using the voter model, focusing on the density of discordant edges and their asymptotic behavior as the graph size increases.

Contribution

It extends analysis of opinion dynamics to directed, heterogeneous graphs using novel coupling techniques and dual processes, broadening understanding beyond regular graphs.

Findings

Asymptotic behavior of discordant edges analyzed

Extension of results to heterogeneous directed graphs

Insights into consensus time and opinion spread

Abstract

We explore the voter model dynamics on a directed random graph model ensemble (digraphs), given by the Directed Configuration Model. The voter model captures the evolution of opinions over time on a graph where each vertex represents an individual holding a binary opinion. Our primary interest lies in the density of discordant edges, defined as the fraction of edges connecting vertices with different opinions, and its asymptotic behavior as the graph size grows to infinity. This analysis provides valuable insights, not only into the consensus time behavior but also into how the process approaches this absorption time on shorter time scales. Our analysis is based on the study of certain annealed random walk processes evolving on out-directed, marked Galton-Watson trees, which describe the locally tree-like nature of the considered random graph model. Additionally, we employ innovative coupling techniques that exploit the classical stochastic dual process of coalescing random walks. We extend existing results on random regular graphs to the more general setting of heterogeneous and directed configurations, highlighting the role of graph topology in the opinion dynamics.