Abstract
<jats:p>This study presents a computational modeling framework for analyzing how risks propagate within multi-agent systems characterized by interdependent political, economic, and local interactions. The approach integrates Bayesian networks, Markov processes, and agent-based modeling to capture both probabilistic dependencies and dynamic feedback loops across multiple system levels. The framework enables the simulation of heterogeneous agents whose adaptive behaviors generate emergent outcomes, allowing sensitivity testing and scenario exploration under varying external shocks. A regional case study regarding the Serbian-Croatian geopolitical hedging is used to validate the methodology, demonstrating how coupled risks (political, energy, and local) interact non-linearly to influence overall system stability. The results highlight critical thresholds at which local perturbations escalate into systemic failures, confirming the framework’s capacity to identify vulnerability patterns and resilience strategies in complex adaptive systems. The proposed integration of probabilistic and agent-based modeling contributes to quantitative risk analytics and provides a transferable tool for decision-support applications across domains involving uncertainty, feedback, and multi-actor interaction.</jats:p>