Chhavi Yadav (UCSD)
Chhavi is a PhD student at UCSD. Her research revolves around various aspects of trustworthy ML, including explainability, auditing and robustness. Chhavi is advised by Prof. Kamalika Chaudhuri. She is supported by Powell and Jacobs School of Engineering fellowships at UCSD. To find more about her, check her website - www.chhaviyadav.org .
Short Abstract: Responsible use of machine learning requires that models be audited for undesirable properties. However, how to do principled auditing in a general setting has remained ill-understood. In this talk, I will propose a formal learning-theoretic framework for auditing. I will give algorithms for auditing linear classifiers for feature sensitivity using label queries as well as different kinds of explanations, and provide performance guarantees. Some of my results illustrate that while counterfactual explanations can be extremely helpful for auditing, anchor explanations may not be as beneficial in the worst case.