Abstract

As machine learning systems become ubiquitous, there is a growing interest in interpretable machine learning—that is, systems that can provide human-interpretable rationale for their predictions and decisions. In this talk, I’ll first give two examples of real healthcare settings—mortality modeling in the ICU and treatment response in major depression—where the ability to interpret learned models is essential, and describe how we built models to meet those needs. Next, I’ll speak about some of the work we are doing to understand interpretability more broadly: what exactly makes a model interpretable? And can we optimize for it? By formalizing these notions, we can hope to identify universals of interpretability and also rigorously compare different kinds of systems for producing algorithmic explanations. Includes joint work with Been Kim, Andrew Ross, Mike Wu, Michael Hughes, Menaka Narayanan, Sam Gershman, Emily Chen, Jeffrey He, Isaac Lage, Roy Perlis, Tom McCoy, Gabe Hope, Leah Weiner, Erik Sudderth, Sonali Parbhoo, Marzyeh Ghassemi, Pete Szolovits, Mornin Feng, Leo Celi, Nicole Brimmer, Tristan Naumann, Rohit Joshi, Anna Rumshisky, and the Berkman Klein Center.