Thomas Burri (University of St. Gallen) and Daniel Trusilo (University of St. Gallen) have posted “Ethical Artificial Intelligence: An Approach to Evaluating Disembodied Autonomous Systems” (In: Rain Liivoja and Ann Valjataga (eds), Autonomous Cyber Capabilities in International Law, forthcoming 2021) on SSRN. Here is the abstract:
Building off our prior work on the practical evaluation of autonomous robotic systems, this chapter discusses how an existing framework can be extended to apply to autonomous cyber systems. It is hoped that such a framework can inform pragmatic discussions of ethical and regulatory norms in a proactive way. Issues raised by autonomous systems in the physical and cyber realms are distinct; however, discussions about the norms and laws governing these two related manifestations of autonomy can and should inform one another. Therefore, this paper emphasizes the factors that distinguish autonomous systems in cyberspace, labeled disembodied autonomous systems, from systems that physically exist in the form of embodied autonomous systems. By highlighting the distinguishing factors of these two forms of autonomy, this paper informs the extension of our assessment tool to software systems, bringing us into the legal and ethical discussions of autonomy in cyberspace.