Augmenting Decision Competence in Healthcare Using AI-based Cognitive Models
Niklas Keller, Mirjam Annina Jenny, Claudia A. Spies, Stefan Michael Herzog · 2020
In many critical decisions, such as medicine, transparency of the underlying decision process is critical. This extends to decision processes that are supported by artificial intelligence. Rather than using a post-hoc explainability approach from explainable AI research (SHAP or LIME), we develop and test an intrinsically transparent and intuitively interpretable model developed from cognitive science, fast-and-frugal trees, in a comparative analysis with state-of-the-art machine learning models. The resultant decision support can be easily implemented as laminated pocket card, augmenting the decision competence of physicians rather than replacing it.