Abstract
This article presents a scoping review of the relevant research discussing the ethics of medical artificial intelligence (AI). Relevant moral and human values can inform the design of ethical medical AI. The value sensitive design (VSD) framework offers a method by which to do this. But much conceptual work must be done in order to apply the VSD framework to medical AI. The goal of this review is to survey existing literature discussing moral and human values (such as responsibility, transparency, and respect for autonomy) in medical AI development. Ultimately, the objective of the review is to advance the VSD methodology for medical AI technologies, in which relevant moral values will inform technology design. Papers were considered eligible if they fulfilled three criteria: (1) provided a discussion of one or multiple moral, ethical, or human values (2) regarding AI or machine learning in a (3) medical, clinical, or health context. We conducted our search on five databases: OVID Embase, OVID Medline, IEEE, Web of Science, and Scopus. After performing our search, we screened title and abstract, then the full text of each paper for relevance using the Covidence platform. After the final sample was determined, we extracted relevant information by coding the papers using NVivo. We found nearly 100 moral values discussed regarding medical AI. In this search we were able to find not only which values are most discussed and how, but also specific, unique, and emerging moral values in the medical AI domain.