Genetic Algorithms for Proportionality Assessment in Military Operations

Abstract

This research proposes the application of Genetic Algorithms (GAs) as an effective computational approach for proportionality assessment in military operations. With this scope, two distinct GA-based models are proposed: one evaluating proportionality excluding psychological damage from collateral damage considerations, and another integrating psychological damage into the collateral damage component of this assessment. Each model encodes the operational assessment rules as binary classification policies, mapping multidimensional states defined by levels of injury, death, object damage, and military advantage to proportional or disproportional decisions. Through evolutionary optimization involving selection, crossover, mutation, and fitness evaluation across generation, the GAs search for classification rules that maximize alignment with expert-defined proportionality judgments. From the evaluation conducted which includes accuracy, fitness progression, population diversity, and confusion matrices, it is seen that the models converge reliably to high-performance solutions, achieving high classification accuracy within the deterministic rule sets. Further, the influence of psychological damage is assessed in relation to the behaviour convergence and classification outcomes. The results show the utility of GAs in automating this military decision-making process as a responsible, transparent, adaptive, and interpretable mechanism for proportionality assessment that is able to incorporate both legal and ethical considerations in operational environments.