“Robots won’t commit war crimes. We just have to program them to follow the laws of war.” This is a rather common response to the concerns surrounding autonomous weapons, and it has even been advanced as a reason that robot soldiers might be less prone to war crimes than human soldiers. But designing such autonomous weapon systems (AWSs) is far easier said than done. True, if we could design and program AWSs that always obeyed the international law of armed conflict (LOAC), then the issues raised in the previous segment of this series — which suggested the need for human direction, monitoring, and control of AWSs — would be completely unfounded. But even if such programming prowess is possible, it seems unlikely to be achieved anytime soon. Instead, we need be prepared for powerful AWS that may not recognize where the lines blur between what is legal and reasonable during combat and what is not.
While the basic LOAC principles seem straightforward at first glance, their application in any given military situation depends heavily on the specific circumstances in which combat takes place. And the difference between legal and illegal acts can be blurry and subjective. It therefore would be difficult to reduce the laws and principles of armed conflict into a definite and programmable form that could be encoded into the AWS and, from which the AWS could consistently make battlefield decisions that comply with the laws of war.