The comparative ethics of artificial-intelligence methods for military applications

Abstract

Concerns about the ethics of the use of artificial intelligence by militaries have insufficiently addressed the differences between the methods (algorithms) that such software provides. These methods are discussed and key differences are identified that affect their ethical military use, most notably for lethal autonomous systems. Possible mitigations of ethical problems are discussed such as sharing decision-making with humans, better testing of the software, providing explanations of what is being done, looking for biases, and putting explicit ethics into the software. The best mitigation in many cases is explaining reasoning and calculations to aid transparency.