About this Event
141 East College Avenue, Decatur, GA 30030
Title: “Normative Concepts for Artificial Moral Agents”
Abstract: Currently existing artificial intelligence (AI) systems are impressive, as they can reliably perform various types of tasks, and in many cases they can perform such tasks just as competently as humans can or even outperform us. Now this might not be too surprising when it comes to certain kinds of tasks, but recent studies of large language models (LLMs) suggest that some of these models might even outperform humans at tasks involving ethical reasoning, including in complicated scenarios in which the ethical answers are not obvious. In this talk, I will explore some potential ramifications of these findings. In particular, I will discuss the possibility that AI systems capable of ethical reasoning (or artificial moral agents (AMAs)) might share our normative concepts, such as our concepts of rightness, fairness, justice, and the like. Even if AMAs are generally capable of providing convincing ethical judgments, I will argue, there is reason to be skeptical about their ethical reasoning capacities if they do not actually share any of our normative concepts. Finally, I will show how this discussion relates to concerns about trustworthy AI and automation bias.
User Activity
No recent activity