Due to copyright issues, an electronic copy of the thesis must be ordered from the faculty. For the faculty to have time to process the order, the order must be received by the faculty at the latest 2 days before the public defence. Orders received later than 2 days before the defence will not be processed. After the public defence, please address any inquiries regarding the thesis to the candidate.
Trial Lecture – time and place
See Trial Lecture.
Adjudication committee
- First opponent: Professor Sabine Salloch, Hannover Medical School,
- Second opponent: Associate professor Leonora Onarheim Bergsjø, Østfold University of Applied Sciences,
- Third member and chair of the evaluation committee: Professor II Anners Lerdal, University of Oslo
Chair of the Defence
Professor II Geir Selbæk, University of Oslo
Principal Supervisor
Professor Bjørn Hofmann, University of Oslo
Summary
This thesis is about responsible explainable artificial intelligence (AI) in medicine, aiming to build a principle-based framework to assess the knowledge that AI offers in the medical field and the ethical implications of AI in medicine.
A key challenge that motivates this work is AI explainability: do we understand why AI systems produce the results that they do, and if not, then is that a problem for medical ethics? The thesis argues that medical ethics should adopt the emerging principle of ‘explicability’ to enable us to make ethical problems visible in AI systems and therefore connect traditional ethical principles to concrete responsibilities. The thesis argues that the principle of explicability, rather than being an unnecessary addition, resembles more established principles in its key focus on connecting abstract theory to clinical practice.
The thesis also proposes a structured tool for evaluating medical AI, called ‘the ethical-epistemic matrix’. While ethical principles are often applied to medical AI, this matrix includes ‘epistemic’ (i.e. knowledge-related) principles, which assess the kinds of knowledge that AI systems can provide. The epistemic principles put forward are: accuracy, consistency, relevance and instrumental efficacy. This thesis also shows how the matrix can be used to explore the implications of medical AI by applying its principles to medical AI systems in general and for the specific case of sleep apnoea detection, which is relevant to the Respire project of which this PhD work is part.
Overall, this thesis advances both the conceptual and methodological foundations for evaluating responsible explainable AI in medicine and provides a practical framework within which to conduct such evaluation. These contributions can support further development of principle- and matrix-based approaches to the evaluation of medical AI, as well as interdisciplinary collaborations aiming to implement responsible explainable medical AI in practice.
Additional information
Contact the research support staff.