Berlin 2024 – scientific programme
Parts | Days | Selection | Search | Updates | Downloads | Help
SOE: Fachverband Physik sozio-ökonomischer Systeme
SOE 20: Focus Session: Statistical Physics of Political Systems
SOE 20.2: Talk
Thursday, March 21, 2024, 15:30–15:50, MA 001
Conflict Classification Using Multinomial Mixture Models and Conflict Avalanches — •Niraj Kushwaha1, Edward Lee1, and Woi Sok oh2 — 1Complexity Science Hub Vienna — 2Princeton University
Armed conflicts are notoriously difficult to systematically characterize and classify[1]. In a recent work, the first problem was tackled using transfer entropy-a measure of information theoretic causality-to group individual conflict events into cohesive cascading structures termed "conflict avalanches"[2]. Our focus now centers on the second problem, wherein leveraging the identified conflict avalanches, we extensively mapped each conflict avalanche to twenty variables commonly linked to armed conflicts in the existing literature. These variables span climatic, socio-economic, demographic, and geographic dimensions. Employing a multi-multinomial mixture model, a novel iteration of the well-established multinomial mixture model, we subjected the conflict avalanches to clustering based on this augmented dataset. The resulting clusters enable us to classify conflicts as compositions of distinct climatic, socioeconomic, demographic, and geographic variables. This systematic classification methodology also identifies crucial underlying determinants for each conflict type and offers insights into the influential factors underpinning each. This innovative classification framework holds promise for advancing our understanding of armed conflicts and improving predictive modeling of armed conflicts. [1] Critical issues in peace and conflict studies: Theory, practice, and pedagogy. Lexington Books, 2011. [2] Kushwaha and Lee, PNAS Nexus, 2023
Keywords: armed conflicts; mixture models; statistical causality; transfer entropy; scaling