Transparency in Medical Artificial Intelligence Systems

International Journal of Semantic Computing, vol. 17, pp. 495–510

Abstract

Many of the artificial intelligence (AI) systems used nowadays have a very high level of accuracy but fail to explain their decisions. This is critical, especially in sensitive areas such as medicine and the health area at large but also for applications of the law, finance etc., where explanations for certain decisions are needed and are often useful and valuable as the decision itself. This paper presents a review of four different methods for creating transparency in AI systems. It also suggests a list of criteria under which circumstances one should use which methods.

Authors 3

1. Lars Quakulinski corresponding Aachen

   RWTH Aachen University

   Affiliation as printed

   RWTH Aachen University, Aachen, Germany

2. Adamantios Koumpis

   University of Cologne · University Hospital Cologne

   Affiliation as printed

   Institute for Biomedical Informatics, Faculty of Medicine and University Hospital Cologne, University of Cologne, Cologne, Germany

3. Oya Beyan

   University of Cologne · University Hospital Cologne · Fraunhofer Institute for Applied Information Technology

   Affiliation as printed

   Fraunhofer Institute for Applied Information Technology (FIT), St. Augustin, Germany

   Institute for Biomedical Informatics, Faculty of Medicine and University Hospital Cologne, University of Cologne, Cologne, Germany