An autonomous system is a machine. For a machine to make adaptive command and control decisions that machine will process data. That data will provide measured pieces of information in order to make decisions and to control behavior. When we discuss Lethal Autonomous Weapon Systems (LAWS) or Unmanned Combat Aerial Vehicles (UCAVs), etc that will be armed, or any systems that make safety critical decisions, they will have to interpret thatvalued information (hopefully in an auditable manner so policies can be traceable to the policy maker and so faulty policies can be corrected if necessary). The Principle of Proportionality talks aboutanticipated military advantage, and uses terms likeexcessive. It assumes that the system can interpretpotential collateral damage. TheCollateral Damage Estimation Methodology may address the weapon system side of the problem, but it does not address the target value in a mathematically explicit manner. While humans might be able to interpret the meaning ofanticipated military advantage andexcessive, they are meaningless to a machine unless the factors can be translated into numbers. So to summarize, for any autonomous system to operate autonomously, the factors controlling the behavior must be mathematically explicit. One might expect that the measured values will be interpreted differently as the operating scenario changes. This adaptivity must also be described in a mathematically explicit manner. There needs to be a forum to formulate the value system for autonomous systems.Back to main menu
Posted:
05/15/2014