Application of AI Health Ethics to Translational Health Science
Muralidharan Anantharaman1, Kathryn Lynn Muyskens1, Michael Dunn1, Annette Braunack-Mayer2, National University Of Singapore 2, New South Wales 1National University Of Singapore Singapore2University of Wollongong New South Wales, Australia
Abstract
AI ethics is a broad and rapidly expanded field. An emerging area of research in this field sets out to identify and address ethical challenges in what we call ‘translational AI health science research’. Reflecting the field of AI ethics more broadly, the literature has tended to favour a top-down approach. On such an approach, one starts with general high-level principles and brings them to bear on specific cases. But this risks trying to shoehorn cases into particular principles or leaving out moral considerations that might be relevant in specific research projects. An alternative approach is to start with specific use-cases, identify and then address distinctive ethical challenges, and to shape and refine principled understandings from the ground up. This panel aims to pursue the second approach through three use-cases. The first use-case examines issues pertaining to meaningful bias in AI with reference to different rates of detection of diabetic retinopathy from retinal scans across groups. The second case examines AI in medical imaging for spinal stenosis and raises questions about the appropriateness of automation and the insistence on keeping ‘humans in-the-loop’. The third case is about the use of AI for triage in emergency situations and covers considerations of equipoise in a technology-mediated and high-risk setting. Together, these cases illuminate many otherwise underexplored challenges of conducting AI-focused research in the clinic.
Biography
All Bios to come