[Paper Review] Noisy bounded confidence models for opinion dynamics: the effect of boundary conditions on phase transitions

Benjamin D. Goddard, Beth Gooding+2 more|arXiv (Cornell University)|Sep 7, 2020

Opinion Dynamics and Social InfluencePhysics and Astronomy82 references37 citations

TL;DR

This paper unifies SDE and PDE bounded confidence models of opinion dynamics under three boundary conditions and analyzes how boundary choices affect phase transitions and convergence, highlighting that no-flux boundaries best reproduce HK-like mechanisms.

ABSTRACT

We study SDE and PDE models for opinion dynamics under bounded confidence, for a range of different boundary conditions, with and without the inclusion of a radical population. We perform exhaustive numerical studies with pseudospectral methods to determine the effects of the boundary conditions, suggesting that the no-flux case most faithfully reproduce the underlying mechanisms in the associated deterministic models of Hegselmann and Krause. We also compare the SDE and PDE models, and use tools from analysis to study phase transitions, including a systematic description of an order parameter.

Motivation & Objective

• Motivate bounded confidence models for continuous opinions under noise and boundary effects.
• Compare SDE and PDE mean-field limits of HK-type dynamics with varying boundary conditions.
• Assess how boundaries influence clustering, consensus, and phase transitions in the presence of radicals.
• Investigate the role of radicals in modifying opinion distributions and cluster formation.

Proposed method

• Formulate SDE and corresponding PDE (Fokker-Planck) models for bounded confidence with noise.
• Implement three boundary conditions: periodic, no-flux, and even 2-periodic on [0,1].
• Incorporate radicals as fixed-opinion agents affecting others via an external potential.
• Define and compute an order parameter to quantify clustering from discrete and continuum perspectives.
• Use pseudospectral numerical methods (Fourier and Chebyshev) to simulate dynamics and study convergence to equilibrium.

Experimental results

Research questions

• RQ1 How do different boundary conditions (periodic, no-flux, even 2-periodic) affect phase transitions and cluster formation in noisy bounded confidence models?
• RQ2 What is the impact of radicals on opinion distributions and the resulting phase behavior under various boundaries?
• RQ3 How do SDE and PDE formulations compare in predicting convergence to equilibrium and transient dynamics?
• RQ4 Which boundary condition best preserves the mechanisms of the original deterministic HK models?

Key findings

• No-flux boundary conditions most faithfully reproduce the underlying mechanisms of the deterministic HK models.
• Radicals can significantly alter clustering behavior, affecting the number and location of opinion clusters.
• SDE and PDE models show corresponding but nuanced agreements in dynamics and equilibrium properties.
• Boundary choice crucially changes the stationary states and convergence paths, with periodic boundaries conflating extreme opinions.
• The order parameter provides a measurable handle on cluster formation and helps characterize phase transitions.