Modeling Interstate War Combat Deaths

Abstract

A prolonged campaign of peaceful interstate competition is an ideal strategic application of artificial intelligence. Monte Carlo simulation, based on validated war analytics, must be at the heart of this capability. Otherwise the system will not know how to assess the potential consequences of failed solutions, chief among them combat fatalities resulting from interstate war. Although the power law has been used since 1960 to model the statistical distribution of deaths resulting from violent conflict, it is not a valid candidate for use in Monte Carlo simulation because it is mathematically divergent for the case of interstate war. Probing Correlates of War Project data, investigators found that combat fatalities in interstate war follow log-gamma or log-normal distributions, depending on whether a state is attacking or defending. Both distributions are valid for use in Monte Carlo simulations. Moreover, they are strong quantitative evidence that war should be modeled as a zero-sum, non-cooperative, high-risk game.