As machine learning models become more integrated into clinical care, how can we update them without violating user expectations? We proposed a new rank-based compatibility measure and loss function to develop clinical AI that better aligns with physician mental models. High rank-based compatibility is not guaranteed but can be achieved through optimization, our approach yields updated models that better meet user expectations, promoting clinician-model team performance.
Presenter:
Dr. Erkin Ötleş is a distinguished Medical Scientist Training Program Fellow at the University of Michigan, specializing in the nexus of AI and medicine. With a robust background in health IT, particularly during his impactful tenure at Epic, he has significantly advanced AI integration into electronic medical records. Erkin’s dedication extends beyond research; he is actively shaping the future of healthcare with his work at the forefront of clinical informatics and AI-driven solutions. He holds a Master's of Engineering from the University of Wisconsin and aspires to further blend clinical practice with technological innovation post his MD-PhD training.
Dr. Erkin Otles Presentation