A New Model for Bias-Generating Agent-Based Simulation and Its Application to Election Systems: Allowing Agents to Make Mistakes for a Reason

Abstract

Several studies have proposed that vote tampering based on heuristic algorithms can manipulate voters’ votes. It can be found from the analysis of the poll results of the 2016 US election that the frequency of “Trump won,” which is generally considered a black swan phenomenon, is not low and even reached 16.8%. However, many models are unable to restore the generation of such a high frequency of black swan phenomena. In this study, the black swan phenomenon is successfully reproduced using a bias-generating agent-based election system model. By adjusting the tampering method, the frequency of the black swan phenomenon will change from 5% to 15%. From the simulation results, it can be observed that one of the possible causes of the black swan phenomenon is the tampering of the voting results, which leads to more biased voters, thus increasing the frequency of the winning elections. This study proposes that to obtain more realistic simulation results, it is necessary to introduce more realistic perceptual models for agents, rather than relying solely on random functions. Allowing agents to make mistakes for a reason should be an integral part of multi-agent-based simulation in the field of pairwise human simulation.