Values, Proportionality, and Uncertainty in Military Autonomous Devices

This position paper presents a discussion on the problem of implementing the rules of International Humanitarian Law in AI-driven military autonomous devices. Fulfilling the proportionality test is one of its key requirements. This test, in order to exclude attacks causing excessive collateral damage, requires comparison of two values: anticipated military advantage and expected incidental harm. In this paper we introduce a discussion of how to take into consideration the problem of uncertainty inherent to all military attacks, and how to combine it with the evaluation process.