A Survey on Ethical Principles of AI and Implementations

Abstract

AI has powerful capabilities in prediction, automation, planning, targeting, and personalisation. Generally, it is assumed that AI can enable machines to exhibit human-like intelligence, and is claimed to benefit to different areas of our lives. Since AI is fueled by data and is a distinct form of autonomous and self-learning agency, we are seeing increasing ethical concerns related to AI uses. In order to mitigate various ethical concerns, national and international organisations including governmental organisations, private sectors as well as research institutes have made extensive efforts by drafting ethical principles of AI, and having active discussions on ethics of AI within and beyond the AI community. This paper investigates these efforts with a focus on the identification of fundamental ethical principles of AI and their implementations. The review found that there is a convergence around limited principles and the most prevalent principles are transparency, justice and fairness, responsibility, non-maleficence, and privacy. The investigation suggests that ethical principles need to be combined with every stages of the AI lifecycle in the implementation to ensure that the AI system is designed, implemented and deployed in an ethical manner. Similar to ethical framework used in biomedical and clinical research, this paper suggests checklist-style questionnaires as benchmarks for the implementation of ethical principles of AI.

From the ground up: developing a practical ethical methodology for integrating AI into industry

“…Most approaches to AI ethics -including industrial AI ethics -are general. The current trend is to work at the level of broad ethical principles, generating sets of principles from expert working groups (Jobin et al, 2019), such as the Ethics Guidelines for Trustworthy AI (EU High-Level Expert Group on Artificial Intelligence, 2019) -HLEG -, or collating such principles together under more general categories (Zhou et al, 2020). The application of these principles is typically developed into 'tools' or 'frameworks,' which are offered not as a series of definite suggestions, but merely as efforts to promote reflection in designers and developers.…”

An exploratory qualitative analysis of AI ethics guidelines

“…The prevalence of the value of transparency has suggested to some that, even if the domain of AI ethics guidelines development encompasses an entangled array of ethical currents or schools of thought, it leans towards a basic deontological logic (Zhou et al , 2020 p. 3012) – that is, in the broadest of brush strokes, an ethical orientation that privileges the notion of an agent’s duty [41]. The findings of this study, however, suggest an alternate reading: AI ethics guidelines display an overarching orientation that is characterized by a logic of utility maximization [42], or one that privileges consequences and thus action or policy that produces the greatest good for the greatest number.…”

Machine Learning and Knowledge Extraction

“…These findings demonstrated important implications in the deployment of AI applications with explanations and fairness. Besides trust and fairness, other AI ethical principles such as accountability also play important roles in trustworthy AI [69]. Our future work will focus on the effects of explanation and introduced fairness on the accountability of AI, as well as the effects of other factors such as application scenarios on the accountability of AI.…”