Validating Simulation-Based Models of Conflict

Abstract

This article discusses the Genesis of War project, in which the author employed the PRISONERS' DILEMMA SIMULATION (PDS) to generate data on the etiology of violent conflict and war as a basis for developing theory relevant to better understanding and dealing with violent ethnic conflicts occurring in post-cold war Europe (e.g., former Yugoslavia). Given the simulation basis of the intended theory construction, there was a need to validate the simulation-generated data before going forward, made more acute by the observation that very little validation of simulation-based data seems to have been done or reported in the literature. Accordingly, a comprehensive validation of the multivariate findings generated by the use of the PDS is presented: an assessment conducted through comparisons with findings from other studies in terms of (a) direction of relationship, (b) magnitude of relationship, (c) ranking of dependent variables by proportion of variation explained, and (d) ranking of independent variables by proportion of variation explained. Across comparisons, the level of agreement between PDS and corresponding (mostly real-world) findings was not only relatively high but also consistent, basically falling within the range of the 75% norm established by Guetzkow's assessment of his INTER-NATION SIMULATION (INS). Such triangulation of findings reinforces the view that some simulations are appropriate for investigating real-world conflict processes.