By Michal Pruski.

I am currently undertaking a mixed-methods project which is looking at barriers and facilitators to the adoption of machine learning in Wales with respect to value-based healthcare – focusing on the potential application of artificial intelligence (AI) to patient reported outcome measures (often known as PROMs). The project is in the early stages, but I thought it would be prudent to start thinking about issues of fairness and equity. These are big topics in healthcare in general, and particularly with respect to the use of technologies which have the potential to increase the digital divide in society.  Writing this paper helped me to think about some of these issues and potentially anticipate some of the themes that will emerge as the project unfolds.

But there was also a second reason why I wanted to work on this topic. Some time ago I was toying with the idea about justice acting as a focal point for bioethical considerations, because it gets mentioned in various ethics frameworks and so could be used to refocus some discussions. I shelved that project away in its early stages, but it kept nagging me from time to time, and I thought that finally I had a good reason to take if off the shelve and unpack it. Nothing like killing two birds with one stone.

Finally, it seems to me that a lot of ethical issues related to healthcare AI have not been explored, but at most merely acknowledged to exist. As part of my higher specialist scientist training in health informatics I found that a lot of scholars are keen on mentioning ethical issues with respect to healthcare AI, but that these issues often remain poorly explored. It is not particularly useful to note that fairness is an issue if you are not going to start thinking what you can do about it. While battles regarding e.g. the superiority of equity over equality (or vice versa) can occasionally be quite intense, so to say, and there are a lot to concepts that need to be addressed, it is the job of the ethicist to ask tricky questions and keep the conversation going. The prospect that there might be a conflict between notions of effectiveness and fairness is not nice, but as we hope to apply more and more digital health technologies to help with patient care, it is important that we consider such problems before implementation. Better not to cause fires, than to run around trying to extinguish them.

Paper title: What does it mean for a clinical AI to be just: conflicts between local fairness and being fit-for-purpose?

Author: Michal Pruski

Affiliations: School of Health Sciences, University of Manchester, UK & Department of Medical Physics and Clinical Engineering, Cardiff and Vale University Health Board, UK

Competing interests: None to declare

Social media accounts of post author: @michal_pruski