Opinion dynamics in two dimensions: domain coarsening leads to stable bi-polarization and anomalous scaling exponents

TL;DR

This paper models opinion dynamics on a 2D lattice, revealing how domain coarsening leads to stable bi-polarization or consensus, with scaling laws influenced by persuasion and compromise probabilities.

Contribution

It introduces a 2D opinion model with persuasion and compromise, analyzing coarsening dynamics, polarization stability, and scaling laws for consensus time.

Findings

Bi-polarization is stable in certain parameter regimes.

Consensus time scales with population size and persuasion-compromise ratio.

Stripe states contribute to long-term polarization stability.

Abstract

We study an opinion dynamics model that explores the competition between persuasion and compromise in a population of agents with nearest-neighbor interactions on a two-dimensional square lattice. Each agent can hold either a positive or a negative opinion orientation, and can have two levels of intensity --moderate and extremist. When two interacting agents have the same orientation become extremists with persuasion probability $p$, while if they have opposite orientations become moderate with compromise probability $q$. These updating rules lead to the formation of same-opinion domains with a coarsening dynamics that depends on the ratio $r=p/q$. The population initially evolves to a centralized state for small $r$, where domains are composed by moderate agents and coarsening is without surface tension, and to a bi-polarized state for large $r$, where domains are formed by extremist agents and coarsening is driven by curvature. Consensus in an extreme opinion is finally reached in a time that scales with the population size $N$ and $r$ as $\tau \simeq r^{-1} \ln N$ for small $r$ and as $\tau \sim r^2 N^{1.64}$ for large $r$. Bi-polarization could be quite stable when the system falls into a striped state where agents organize into single-opinion horizontal, vertical or diagonal bands. An analysis of the stripes dynamics towards consensus allows to obtain an approximate expression for $\tau$ which shows that the exponent $1.64$ is a result of the diffusion of the stripe interfaces combined with their roughness properties.