A Minimalist Physics-Informed Model for Predicting Extreme Conflict Fatalities

TL;DR

This paper introduces a minimalist, physics-informed Bayesian model to predict extreme conflict fatalities at the country level, emphasizing the role of entropy and energy states in conflict severity.

Contribution

The study proposes a novel, theoretically grounded approach combining physics principles with Bayesian modeling to improve predictions of extreme conflict fatalities.

Findings

The model effectively predicts extreme conflict events.

Physics-informed measures outperform traditional methods.

Entropy and energy states are key indicators of conflict severity.

Abstract

The complexity of armed conflicts is expressed in the number of fatalities that may span several orders of magnitude. This study presents a minimalist, physics-informed approach to estimating the likelihood of extreme conflict fatalities at the country level of analysis using Bayesian modeling and energy-based dynamics. Leveraging the Boltzmann distribution to construct a Dirichlet prior, we formulate a predictive measure that captures the underlying entropy and energy states of conflict severity. By analyzing a dataset of 112 countries in conflict, we support the predictive power of the proposed measure. The findings suggest that extreme conflict events may be better understood through a minimal but theoretically grounded approach.