Assessing and Mitigating Unfairness in AI Systems

Explore the critical field of fairness in AI systems through a comprehensive tutorial focused on assessing and mitigating unfairness in the context of the U.S. healthcare system. Dive into a scenario involving patient health risk modeling with demonstrated racial disparities. Gain hands-on experience using Jupyter notebooks and the Fairlearn library to assess ML model performance disparities across racial groups and implement various algorithmic techniques for mitigation. Learn to explore, document, and communicate fairness issues effectively using resources like datasheets for datasets and model cards. Designed for participants with intermediate Python skills and familiarity with Scikit-Learn, this 2-hour 38-minute PyCon US tutorial combines instructional content with practical demonstrations to address the negative impacts of AI systems on historically underserved and marginalized communities.