Systems that are capable of producing change in the environment require particular attention to the ethical ramifications of their behavior. The determination and mitigation of the ethical concerns of such systems has to date been charged to their designers and has largely been accomplished by simply preventing systems from engaging in ethically unacceptable behavior in a predetermined, ad hoc manner, which can unnecessarily constrain the set of possible behaviors. Although this may have been considered "best practice" in the past, the coupling of computational intelligence to such systems is likely to provide better options. This is especially true of autonomous systems, which not only produce change in the environment but are capable of monitoring this environment to determine the effect of their actions as well as what the next action should be. Ethical questions concerning the behavior of such complex and dynamic systems are likely to exceed the grasp of their designers and elude simple static solutions. Autonomous systems will therefore need tools and methodologies to help codify ethical principles pertinent to their behavior, principles which form the basis for the autonomous selection and justification of ethically preferable actions.
In this research, the PI and his team will develop and implement a general methodology for the discovery of ethical principles, abstracted from their previous work, which incrementally constructs representations that characterize ethical dilemmas and discovers decision principles necessary to resolve them. They will use this system to codify representation schemes and principles of ethical decision-making in domains that are of particular significance to autonomous systems in their interaction with human beings. And they will evaluate the discovered representations and principles through independent consultation. The PI expects project outcomes will provide evidence that ethical principles and decision-making can be computed and function effectively in domains where machines are likely to interact with human beings.
Broader Impacts: This research will lay the foundations for providing autonomous systems across multiple domains with the principles required to choose the most ethically correct actions and justify these choices. Thus, the work should alleviate concerns with autonomous systems and bolster society's support for their development in a wide range of domains. An important by-product of the research is that breakthroughs in ethical theory are likely to result as representations and principles for resolving ethical dilemmas are discovered.
