Zealots in multi-state noisy voter models

TL;DR

This paper explores how zealots influence the complex phase behavior of multi-opinion noisy voter models on complete graphs, revealing up to six distinct stationary states and the impact of partial zealotry.

Contribution

It extends the analysis of zealots' effects from two-opinion models to multi-opinion scenarios with M>2, uncovering richer phase structures and dynamics.

Findings

Up to six different stationary states identified.

Zealots can eliminate some phases but not all.

Partial zealotry modeled as fractional zealots affects phase behavior.

Abstract

The noisy voter model is a stylised representation of opinion dynamics. Individuals copy opinions from other individuals, and are subject to spontaneous state changes. In the case of two opinion states this model is known to have a noise-driven transition between a unimodal phase, in which both opinions are present, and a bimodal phase in which one of the opinions dominates. The presence of zealots can remove the unimodal and bimodal phases in the model with two opinion states. Here, we study the effects of zealots in noisy voter models with M>2 opinion states on complete interaction graphs. We find that the phase behaviour diversifies, with up to six possible qualitatively different types of stationary states. The presence of zealots removes some of these phases, but not all. We analyse situations in which zealots affect the entire population, or only a fraction of agents, and show that this situation corresponds to a single-community model with a fractional number of zealots, further enriching the phase diagram. Our study is conducted analytically based on effective birth-death dynamics for the number of individuals holding a given opinion. Results are confirmed in numerical simulations.