Artificial intelligence in medicine: trust it or (merely) rely on it?

Abstract

Problem: Trust and trustworthiness are usually highly valued in modern medicine. These phenomena are seen as components of a good patient–doctor relationship (P-D-R). In the field of artificial intelligence (AI), they are important regulatory reference points. However, it is not entirely clear what we mean when we talk about trust or trustworthiness in these areas. Argumentation: It therefore seems worthwhile to explore the meaning of these terms and the implications for debates in medical ethics. It is argued here that trust involves a complex, noncontrolling, interpersonal attitude implying that predicating trust on nonpersonal entities—such as AI—is a kind of category mistake. In the context of evaluating AI, reliability is a much more appropriate candidate. Conclusion: In turn, the reliability of AI should be proven and controlled in according to the idea of technovigilance, especially with a system view, which implies the control of humans in the loop, i.e., doctors. However, this technovigilance-related control of doctors does not undermine a good P‑D‑R in which trust can have a good place.