Most research in medical AI never makes it to the clinic. We aim to create more
clinically useful AI and increase technology acceptance among clinicians by
establishing Human-AI collaboration as a target that can be optimized similarly to
predictive performance. In terms of explainable AI, this defines a shift from
researching what we can explain to also researching how we explain it well.
Implementation and validation will take place in medical simulation centers, in which
Human-AI collaboration can be observed and manipulated in a controlled way.