The ethics of robot warriors
AMany people agree that Isaac Asimov's Three Laws of Robotics set the basic, if somewhat self-preserving, guidelines on how our largely autonomous metallic creations should interact with us fleshier beings: Don't hurt humans, obey humans unless one human orders you to hurt some other human, and don't hurt yourself except to save humans from danger.
Pretty clear-cut, right? But what about robots in combat? What about robots we've built to wipe out enemy combatants, terrorists and other threats to our freedom? In short, how do we program ethical behavior into a robot designed to engage in combat with a human? With our increasing reliance on unmanned aerial vehicles and iRobot's surveillance and bomb-disarming bots, the question may not be so speculative in the years to come.
Researcher Ronald Arkin at the Georgia Institute of Technology's Mobile Robot Laboratory has grappled with the issue in a new paper, 'Governing Lethal Behavior: Embedding Ethics in a Hybrid Deliberative/Reactive Robot Architecture' (GCN.com/955).
Examinations of the ethics of war are nothing new. Arkin reviewed the laws of combat through the ages, noting that in human-to-human combat, it is not proper to attack civilians or even soldiers who have laid down their weapons in surrender.
But Arkin also formulated a machine-ready algorithm for ethical behavior, a tricky undertaking given that even humans find all-too-many potential actions in combat ethically murky at best.
Arkin's approach was to first 'describe the set of all possible behaviors capable of generating a discrete lethal response'that an autonomous robot can undertake.'
Then he formulated a set of ethical constraints ' based on the Geneva Conventions and other largely agreed-upon ethical norms for war ' and applied them to this set of lethal behaviors.
The resulting set of ethically lethal actions could then be implemented through a number of architectures and even a pseudocode that Arkin offered.
This is, of course, is a much-simplified and abstract approach, and it took Arkin nearly 100 pages to formalize such inherently loose concepts as return fire, ambush and other tried-and-true military tactics. Arkin said his work is only beginning, but he is optimistic about future developments.