Consensus in Models for Opinion Dynamics with Generalized-Bias

TL;DR

This paper extends the DeGroot opinion model to include generalized cognitive biases, demonstrating that consensus can still be achieved in strongly connected social influence networks despite diverse biases.

Contribution

Introduces generalized-bias opinion models that incorporate various cognitive biases, broadening the understanding of opinion dynamics beyond traditional models.

Findings

Consensus is achievable under certain conditions despite biases.

Models capture dynamic influences, ingroup favoritism, and out-group hostility.

All agents converge to a consensus in strongly connected influence graphs.

Abstract

Interest is growing in social learning models where users share opinions and adjust their beliefs in response to others. This paper introduces generalized-bias opinion models, an extension of the DeGroot model, that captures a broader range of cognitive biases. These models can capture, among others, dynamic (changing) influences as well as ingroup favoritism and out-group hostility, a bias where agents may react differently to opinions from members of their own group compared to those from outside. The reactions are formalized as arbitrary functions that depend, not only on opinion difference, but also on the particular opinions of the individuals interacting. Under certain reasonable conditions, all agents (despite their biases) will converge to a consensus if the influence graph is strongly connected, as in the original DeGroot model. The proposed approach combines different biases, providing deeper insights into the mechanics of opinion dynamics and influence within social networks.