Dynamics of bounded confidence opinion in heterogeneous social networks: concord against partial antagonism

TL;DR

This paper introduces a mixed opinion dynamics model incorporating two psychological types of agents, analyzing how their interactions influence social consensus, polarization, and opinion fragmentation across different network topologies.

Contribution

It proposes the Mixed model that accounts for both concord and partial antagonism agents, extending the Deffuant-Weisbuch model to heterogeneous social groups.

Findings

Opinion formation is largely topology-independent.

Mixed societies exhibit unique bifurcation patterns.

Initial opinion uncertainty significantly impacts collective opinion shifts.

Abstract

Bounded confidence models of opinion dynamics have been actively studied in recent years, in particular, opinion formation and extremism propagation along with other aspects of social dynamics. In this work, after an analysis of limitations of the Deffuant-Weisbuch (DW) bounded confidence, relative agreement model, we propose the Mixed model that takes into account two psychological types of individuals. Concord agents (C-agents) are friendly people; they interact in a way that their opinions get closer always. Agents of the other psychological type show partial antagonism in their interaction (PA-agents). Opinion dynamics in heterogeneous social groups, consisting of agents of the two types, was studied on different social networks. Limit cases of the mixed model, pure C- and PA-societies, were also studied. We found that group opinion formation is, qualitatively, almost independent of the topology of networks used in this work. Opinion fragmentation, polarization and consensus are observed in the mixed model at different proportions of PA- and C-agents, depending on the value of initial opinion tolerance of agents. As for the opinion formation and arising of "dissidents", the opinion dynamics of the C-agents society was found to be similar to that of the DW model, except for the rate of opinion convergence. Nevertheless, mixed societies showed dynamics and bifurcation patterns notably different to those of the DW model. The influence of biased initial conditions over opinion formation in heterogeneous social groups was also studied versus the initial value of opinion uncertainty, varying the proportion of the PA- to C-agents. Bifurcation diagrams showed impressive evolution of collective opinion, in particular, radical change of left to right consensus or vice versa at an opinion uncertainty value equal to 0.7 in the model with the PA/C mixture of population near 50/50.