UKRI AI CDT in Dependable and Deployable AI for Robotics (CDT-D2AIR)

Fleur Conway

Project
Addressing the AI Deployment Gap in Critical Care

Summary
I aim to assess the deployment gap of using artificial intelligence as a decision support tool in intensive care units. This comprises three areas. The first is to assess the utility of such models in both improving patient outcomes and meeting clinician needs. The second is to explore and evaluate different explainable modalities - such as building inherently interpretable models or using post-hoc methods. Finally, to address the human factor in interacting with such explanations - for example liability, appropriate reliance, and cognitive burden. During this process I hope to build and test models and explainability interfaces, informed by co-design with clinicians, patients and regulators.