Validation Methodology for Agent-based Simulations

摘要

Agent-based simulations are desirable because they simulate interesting emergent aggregate behavior using modular reasoning applied to local conditions. The goals of simulation testing, part of the process of validation, is to determine if the simulation produces predicted and controlled results -- hence, we construct agent-based simulations predisposed to fail simulation testing procedures. In this work, we will describe our thinking on the alterations to the validation and testing process that are necessary to accommodate agent-based simulations.