Understanding medical black boxes: a philosophical analysis of artificial intelligence explainability


  • Dr Rune Nyrup

    University of Cambridge

Project summary

While modern artificial intelligence (AI) systems can display impressive accuracy in medical tasks, it can be difficult to fully explain how they work or why they make individual decisions. This project will study the ethical challenges arising from the application of black box AI systems in medical contexts.

I will apply a pragmatic account of AI explainability, which focuses on the kind of understanding patients and physicians require to take relevant actions for a contextually relevant purpose. This makes AI explainability dependent on context as different kinds of understanding are relevant in different contexts and for different users.

This project will investigate the challenges AI systems raise for understanding in three specific medical contexts: testing whether drugs are effective; doctors’ clinical reasoning and patients giving meaningful informed consent.