Regensburg 2025 – scientific programme
Parts | Days | Selection | Search | Updates | Downloads | Help
SOE: Fachverband Physik sozio-ökonomischer Systeme
SOE 6: Political Systems and Conflicts
SOE 6.2: Talk
Tuesday, March 18, 2025, 14:30–14:45, H45
Knowing armed conflict type hurts prediction — •Niraj Kushwaha1, Edward D. Lee1, and Woi Sok Oh2 — 1Complexity Science Hub, Austria — 2Princeton University, USA
Moving beyond heuristic classifications of armed conflicts such as local, civil, interstate wars etc. to systematic categorization is useful but challenging. Using information-theoretic techniques we generate chains of related conflict events from the high-resolution Armed Conflict & Location Event Dataset and integrate them with other datasets spanning climate, geography, infrastructure, economics, raw demographics, and composite demographics. Using an unsupervised clustering algorithm based on multinomial mixture, we discover that three conflict archetypes exist; “major unrest,” “local conflict,” and “sporadic & spillover events.” Major unrest predominantly occurs in densely populated areas with good infrastructure and flat, riparian geography. Local conflicts arise in mid-populated regions with diverse socio-economic and geographical features. Sporadic and spillover conflicts are small, occurring in sparsely populated areas with poor infrastructure and economic conditions. The three types stratify into a hierarchy of factors, revealing a quantitative taxonomy that highlights population, infrastructure, and economics as the most discriminative variables. Surprisingly, we find that knowing the type negatively impacts predictability of conflict intensity such as fatalities, conflict duration, and other measures of size. Hence, we develop an empirical and bottom-up approach that identifies conflict types but also cautions us about the limited utility of public data sets for conflict prediction.
Keywords: Armed Conflict; Clustering; Prediction; Conflict Avalanche; Multiscale