Loading Events
  • This event has passed.

« All Events

AI For Clinical Diagnostic Decision Making: Can Explainability be a Backstop Against Biased AI?

April 2 @ 12:00 pm 1:00 pm

Part of the Planning Workshop #1

Sarah Jabbour
Ph.D. Candidate, Computer Science and Engineering
University of Michigan

Abstract: There is a critical need for artificial intelligence (AI) in healthcare, and its potential warrants genuine enthusiasm. For example, AI models could help clinicians make faster and more accurate diagnostic decisions, resulting in improved patient outcomes and minimizing treatment delays. However, the safe adoption of these tools will require careful consideration during both model development and model deployment. For instance, AI tools are susceptible to bias found in the data we train them on, and could negatively impact clinical diagnostic decisions if not integrated carefully into clinical care.

In this talk, I will discuss my efforts in this area. I’ll first show how such biased behavior arises during model development and our proposed approaches to improve model performance. I’ll then present a clinician-AI interaction study in which we measured the impact of biased AI and AI explanations on clinicians diagnostic and treatment decisions. 
Bio: Sarah Jabbour is a doctoral student in Computer Science and Engineering (CSE) at the University of Michigan, where she is advised by Professors Jenna Wiens and David Fouhey. Her primary research interests lie at the intersection of machine learning, human-AI interaction, and healthcare, with a particular focus on the development of AI models for aiding clinicians in diagnosis and treatment decisions.