November 19, 2020 - 12:00pm to 1:00pm
James Zou: "Finding and reducing gender stereotypes in AI systems"
Abstract: Professor Zou will present a framework to audit AI systems to characterize gender and ethnic stereotypes that are embedded in the algorithm due to biased training data. Then he will discuss some practical approaches for reducing such algorithmic stereotypes. He will also show some examples of how we can turn these algorithmic biases into new tools for studying history.
James Zou is Assistant Professor of Biomedical Data Science.