Browse Books

Reviewer: George A. Bekey

Robots have evolved rapidly from science fiction and university laboratories to becoming commonplace in industry, education, entertainment, and the military. As their capabilities and ability to operate autonomously have grown, so has concern about possible dangers from their deployment, especially in military applications where they may be equipped with lethal weapons. At long last, questions on the ethical use of robots are being raised. Until the year 2000, the only significant voice on this matter was that of Bill Joy, the former Chief Scientist of Sun Microsystems. Joy raised questions about the potential dangers of autonomous robots, nanotechnology, and genetic engineering [1]. More recently, Wallach and Allen raised a series of ethical issues in connection with autonomous robots [2]. Singer's Wired for war [3] and Krishnan's Killer robots [4] are expositions of developments in robot technology that may dramatically change the nature of war. Now, Arkin, a leader in robotics research and applications, has written this remarkable book concerning methods of controlling potentially lethal actions by autonomous robots, based on his long experience with military robotics programs. Arkin's book is based on three fundamental assumptions. First, wars will continue to be waged, as they have from time immemorial. This assumption is clearly based on historical evidence and substantiated by psychological research that documents humanity's simultaneous attraction and aversion to war [5]. Second, it is essential to regulate robots' behavior during wartime to insure that, at the very least, they conform to the laws of war-the internationally accepted codes of conduct for military personnel, as embedded in the Geneva Conventions-and the rules of engagement (ROE)-used by US military services to clarify the laws of war. Third, while it may not be possible to design robots that always behave ethically, they can be built to be more ethical than human soldiers in wartime. These assumptions lead to the notion of "ethically-justified lethal behaviors" on the part of military robots. The major portion of the book is devoted to a formalization of this concept, including the development of software architecture for military robots equipped with lethal weapons, which makes such ethical behaviors possible. There are certainly people who consider ethical killing an oxymoron, but as Arkin points out, the notion of a just war and the international acceptance of the laws of war emphasize limits on the use of lethal weapons. Thus, as discussed in chapter 7, the laws of war require the following: military actions should attempt to limit casualties among noncombatants; wounded enemy soldiers are to be assisted; and the use of lethal force in response to an attack should be proportional to the scale of the attack. These restrictions on the absolute use of lethal force form the basis of the software architecture proposed by Arkin for governing the behavior of military robots. The first three chapters describe current military robots equipped with lethal weapons and the failings of human soldiers that may lead to violations of the laws of war. Chapter 4 is a review of philosophical approaches to the question of ethics, and chapter 5 is a fascinating report on a survey undertaken by Arkin's laboratory to find out what people think about robots equipped with lethal weapons. Chapter 6 develops the formalism used as a basis for designing ethical controllers. Chapter 7 is "Specific Issues for Lethality: What to Represent." Chapter 8 is "Representational Choices: How to Represent Ethics in a Lethal Robot." The second half of the book begins with chapters 9 and 10, which discuss architectural considerations and design options for the design of ethical controllers. Chapter 11 presents recent military actions in which lethal force was used and describes applicable ethical considerations. The concluding chapter presents a prototype implementation of a control architecture that includes a number of carefully designed restraints on the robot system, to ensure that its use of lethal force conforms to the ethical guidelines laid out in the laws of war and ROE. This book is a courageous step in the design of autonomous robotic systems that satisfy a set of ethical constraints. It should be required reading for anyone in the computer science (CS) and engineering communities who is interested in military robots and their deployment. It is easy to condemn the military services for using robots equipped with deadly weapons, but Arkin has taken a much more constructive course of action, by showing us how to control them using ethical guidelines. Online Computing Reviews Service