97 Things About Ethics Everyone in Data Science Should Know

Chapter 92. Foundation for the Inevitable Laws for LAWS

Stephanie Seward

Associate, Booz Allen Hamilton¹

Advances in AI technology have led to justified calls for banning lethal autonomous weapons systems (LAWS). Governments worldwide answer these calls with silence. When countries, methods of governance, and lives are at stake, leaders resort to extreme methods to ensure survival and protect their citizenry. In the future of warfare, the ability of LAWS to assess and engage targets quickly will give a decisive advantage to the country that possesses the most advanced technology and a willingness to use it. In warfare, the technology required to win prevails. Considering these stark realities, what degree of certainty do AI machines require to engage targets autonomously? The following is a methodology that practitioners can use to dictate how much freedom to give autonomous systems in making or informing vital decisions.

Performance Expectation Methodology (PEM)

A model trained on one set of data and tested on a separate set of data has a baseline metric for accuracy that may not be representative of real-world scenarios. The PEM is designed to test LAWS in a manner representative of real-world scenarios and should serve as the baseline for testing requirements that LAWS must pass before they are deployed to make independent decisions. Much as a doctor who achieves excellent ...