At the workshop for the thematic research network Human Intelligence for Artificial Intelligence in Medicine (HI4AI), the EMCL in the person of Marcus Buchwald presented its research in collaboration with the Mannheim Institute for intelligent Systems in Medicine. The event, held on February 19, 2024, showcased innovative approaches to improving healthcare through collaborative projects with an AI focus.

The presentation, titled “Challenges in Learning to Defer to Experts” explored a hybrid decision-making framework in which the integration of machine learning (ML) is used to learn when to defer to a medical expert on top of an AI prediction. The goal is to boost the performance by employing a joint-prediction system of AI and experts in comparison to the stand-alone rater systems of experts and AI respectively. As such, the key goal is to developing algorithms that can intelligently decide when to rely on AI predictions and when to defer to a human expert. The research presented demonstrated the challenges and possibilities of learning to defer by exploring diverse approaches and testing them on a handcrafted dataset - Glomerular Experts dataset, which was curated as part of the research.
 

The 1st TRN HI4AI Workshop can be viewed as a step forward in fostering collaboration between artificial intelligence and human intelligence in the medical field, showing the way for more AI-integrated and effective healthcare solutions.

Author: Marcus Buchwald

Date: 25/02/24