Bayesian Network Enviroment for Agent Based Modeling Proliferation Risk Analysis

Year
2013
Author(s)
Royal Elmore - Texas A&M University Nuclear Security Science and Policy Institute
William S. Charlton - Texas A&M University Nuclear Security Science and Policy Institute
Abstract
Agent Based Modeling offers capabilities for assessing intelligent and innovative nuclear proliferation adversaries and adaptive counter proliferation entities. With agent based modeling individual agent entities possess certain factors they seek to optimize when interacting with other entities. Development of a nuclear proliferation framework was needed, in which the agents could operate to further their proliferation objectives. Bayesian analysis was undertaken to develop the needed proliferation network. The Netica Application Programmer Interfaces-C modules were used within the Microsoft Visual Studio 2012 C++ program to develop a Bayesian proliferation network. Agents are introduced within the Bayesian network in three broad categories: neutral, proliferating, and defensive agents. Proliferating agent objectives for considering pathways include technical limitations, relative economic cost, time, and difficulty of outside detection. Repeated simulations with small perturbations demonstrate the impact of small proliferation network perturbations. Alterations in the proliferation network affect the proliferating agent’s ability to prioritize a particular objective in obtaining a desired nuclear posture. The connections and relative capabilities of neutral, proliferating, and defensive agents within a proliferation simulation can lead to many outcomes. Various affinities for different agent interactions could then lead to new or restricted proliferation options.