Agent mental models and Bayesian rules as a tool to create opinion dynamics models

Abstract

Traditional opinion dynamics models are simple and yet, enough to explore the consequences in basic scenarios. But, to better describe problems such as polarization and extremism, we might need to include details about human biases and other cognitive characteristics. In this paper, I explain how we can describe and use mental models and assumptions of the agents using Bayesian-inspired model building. The relationship between human rationality and Bayesian methods will be explored, and we will see that Bayesian ideas can indeed be used to explain how humans reason. We will see how to use Bayesian-inspired rules using the simplest version of the Continuous Opinions and Discrete Actions (CODA) model. From that, we will explore how we can obtain update rules that include human behavioral characteristics such as confirmation bias, motivated reasoning, or our tendency to change opinions much less than we should.

A Reputation Game Simulation: Emergent Social Phenomena from Information Theory

“…Here, instead a top down approach is chosen, in which a certain amount of rationality of the agents' minds is postulated in the spirit of Bayesian models of cognition. [70,71] Also, we assume that rationality tries to align with Bayesian reasoning (or probabilistic logic) [30,39] and follows information theoretical principles. [31,38,62] Bayesian frameworks are used in many works on computational psychology and cognition for example see ref.…”