Machine Learning and Generative Models for Conflict Prediction and Understanding

• In Event: New Methods in the Study of Conflict Dynamics

Abstract

There has been an increasing focus on conflict predicting and forecasting across international relations. However, researchers have been forced to choose between either simple, low performing conceptually-justified models or flexible, high performing black box algorithms. Here we suggest that black box algorithms can be used to improve not only conflict forecasting metrics, but also human understanding (and the formalization) of conflict processes. Using PRIO grid-cell level data, we are able to show how the relative improvements in prediction from a black box model can illuminate missing pieces from existing analytical generative models; suggesting new parameterizations. In particular, we learn that a Bayesian mixture model can close the performance gap with ML models. Our results highlight how advances in machine learning can accelerate social science research on conflict prediction.

Authors

• Michael P. Colaresi, University of Pittsburgh

• Kevin T Greene, University of Pittsburgh