Utilizing User Stories to Bring AI Ethics into Practice in Software Engineering

Abstract

AI ethics is a research area characterized by a prominent gap between research and practice. With most studies in the area being conceptual in nature or focused on technical ML (Machine Learning) solutions, the link between AI (Artificial Intelligence) ethics and SE (Software Engineering) practice remains thin. Establishing this link, we argue, is vital going forward. While conceptual discussion is required to define AI ethics, much progress has already been made in this regard. Similarly, though technical ML solutions are also required for practical implementation, ML systems are ultimately still software, and thus SE cannot be forgotten. In this paper, we propose one way of bringing AI ethics closer to conventional SE practice: utilizing user stories to implement AI ethics by means of Ethical User Stories (EUS). EUS can be used to formulate both functional and non-functional requirements, although an ethical framework is required produce them. By treating AI ethics as a part of the development process in this fashion, as opposed to a separate task, it can ideally become a part of SE for ML systems.