Modeling Polarization in Mass Populations using Agent-Based Modeling & Novel Opinion Dynamics

By Justin Mittereder, Robert Carroll, Brandon Frulla

Faculty Mentor: Professor Stephen Davies

The 21st century has opened doors for large-scale simulations that were previously impossible due to computing power restrictions. Now, we are able to create large simulations of heterogeneous “agents”. These simulations allow researchers to discover what set of simple rules (behavioral rules of each agent) are sufficient to produce a particular phenomenon on the societal level. The particular phenomenon we hope to observe is political polarization. The results of this research will provide insights into how polarization arises, and how it may be prevented from escalating further. In our simulation, agents will have a predetermined number of different opinions that are assigned randomly from 0-1. At each step of the simulation, agents will choose another neighboring agent at random for an interaction. Then, they will look to see if they agree closely on a random issue. If the opinions of both agents are within a predetermined comparison threshold, then we will take the average of the neighbor and the agent’s opinion on another different issue and set that as the agent’s new opinion on that given issue. Throughout the life of the simulation, we will measure a number of different variables such as the average assortativity across all issues, the average opinion variance, the average persuasions per agent, and the number of opinion clusters for each issue. We will examine the data from many simulation runs to look for emergent behavior across all agents in the simulation. Our goal is to manipulate the parameters of the model in such a way that sheds light on how polarization develops in a society. Some of the tools used in this project are Mesa, Networkx, Python and Dash.

Transcript_Justin-Mittereder

Leave a Reply

Your email address will not be published. Required fields are marked *