Ethical Adaptation: Exploring the Use of Adaptive Autonomy in the Design of Ethical AI Teammates in Healthcare

Abstract

Modern advancements in AI technologies have allowed for their more seamless integration within society, including as full-fledged teammates tasked with making and executing independent decisions. This increased integration amplifies the burden on designers to determine if and when AI is capable of making such decisions when confronted with an ethical dilemma. In this mixed-methods study, we conducted a factorial survey (N=200) and interviewed fifteen medical professionals to understand how the principles of medical ethics should affect AI teammate autonomy and behavior. The results of this study enabled the creation of important themes and design recommendations that can guide the design of ethical AI teammates that can appropriately recognize and adapt to ethical dilemmas.