Machine learning for precision medicine: promoting value considerations through perspective-taking hypothetical group design exercises

Abstract

Public concerns over the social and ethical consequences of artificial intelligence (AI) are well established. Despite ongoing efforts to respond, these concerns remain largely unresolved by either regulation or codes of ethics. In response, scholars have advanced ideas about how to better ground ethical awareness in practice. A key element of this grounding is fostering awareness of how one’s actions can affect the welfare of others. We tested the effect of a group design exercise on whether and how AI developers considered the impact of their work on others, using perspective-taking as a “values lever”–a practice that prompts ethical reflection during the design process. We found that hypothetical scenarios set in three different contexts of AI research or building a tool for clinical use encouraged developers to take different perspectives. We specifically used an imagine-self framing to instruct AI developers to think about how they would feel or act in a certain situation. In progressing through the scenarios, developers’ design considerations shifted from methodological and data concerns to thinking about other interest holders, implementation, and social and ethical issues. In particular, a scenario that used the imagine-self framing appeared to lead to greater consideration of the patient perspective, self-awareness of this shift in perspective, and how it might and should affect their future practice. We conclude that a brief group exercise can increase awareness of the impact of design considerations on a broad range of interest holders, and inspire plans for action in future work.

Ähnliche Arbeiten