Ethics for artificial agents

Abstract

Machine ethics is a nascent subfield of computer ethics that focuses on the ethical issues involved in the design of autonomous software agents ("artificial agents"). Chapter 1 of this thesis considers how best to understand the central projects of this new subfield, and reconstructs a prominent theory of how artificial agents ought to be designed. This theory, which I call the "agential theory" of machine ethics, says that artificial agents morally ought to be designed to behave only in ways that would be permissible for a human agent to behave, and that only artificial agents that have been designed in this way are morally permissible for human beings to use. Chapter 2 critically assesses two versions of the agential theory-one that assumes that artificial agents are moral agents, and another that does not. After considering arguments for both versions of the theory, I argue that both versions should be rejected. Chapter 3 sets out and analyzes a case study in machine ethics, focusing on the development of an artificial agent to assist with the planning of a public health social work intervention.