Causal Fairness for Machine Learning
Addressing AI bias through causal fairness, emphasizing the importance of causal effects in measuring and mitigating bias
Our team studies causal fairness, using causal effects to measure and mitigate bias in AI systems. We evaluate models through causal and counterfactual analyses, including scenarios in which protected attributes change. In collaboration with the University of Arkansas, we are extending this work to dynamic and non-iid (non-independent and identically distributed) settings.