Speaker: Yves Saint James Aquino

Artificial intelligence (AI), specifically machine learning (ML), systems hold great potential in improving clinical processes, from diagnosis and screening to targeted therapy. One of the key challenges of implementing ML systems is the regulation of something that evolves over time. While regulation of an evolving product is not entirely new (e.g. medical software), changes in ML are particularly challenging as they tend to be more data-driven, may be independent of programmers, and may not be easily explainable, among others. This paper will tackle this regulatory challenge by framing it as a problem of “algorithmic identity”, drawing on scholarship from metaphysics and philosophy of science. I will discuss four types of problems with respect to the identity of ML systems that are subject to regulatory approval. I will then discuss the ethical and regulatory implications of these problems in the context of healthcare and medical device regulation. Finally, I will discuss a preliminary framework for addressing the problem of algorithmic identity in the regulation of AI/ML in healthcare.