Cole Frank

The through line of my career — first as a macroeconomics researcher, then as an ML engineer, and most recently as a PhD student — is a fascination with the mathematical models we build of the social phenomena around us.

I'm a PhD student at CMU's Heinz College of Information Systems and Public Policy, interested in how machine learning can better serve the public interest. My research has two threads:

1. Holistic evaluations of ML systems — designing better assessments of the fairness and privacy dimensions of models already being deployed in high-stakes domains.
2. New methods for causal inference — further incorporating ML into the traditional econometric toolkit for program evaluation.

Models help explain the world and automate otherwise resource-intensive tasks. But the gaps that inevitably open up between the model and the real world result in real costs for the people affected by those models. As more human decision-making is ceded to models, it becomes ever more important to measure and ameliorate their shortcomings.