Title: Economic Modeling and Machine Learning

Abstract

Machine learning has demonstrated tremendous predictive success in problems ranging from medical diagnosis to which criminal defendants should be released on bail.

Economic modeling, however, is often directed at least in part towards advancing understanding of the behaviors we see. Black-box algorithms that generate substantial improvements in prediction may not (indeed, often do not) generate comparable improvements in insight. Can machine learning algorithms be leveraged as a tool for social science modelers?

This talk will propose methodologies for use of machine learning techniques to complement and improve traditional economic modeling. Specifically, I will discuss work that applies machine learning techniques to (1) evaluate the predictive capabilities of economic models, and (2) identify interpretable extensions of existing models. These methodologies are used to derive new insights into specific economic problems, and may be applied more broadly within the social sciences.

Bio

Annie Liang is an Assistant Professor of Economics at the University of Pennsylvania. She is an economic theorist whose work focuses on information economics, and the application of machine learning techniques for model building and evaluation.

Prior joining UPenn, Annie was a post-doc at Microsoft Research-New England. She received a B.S. in Mathematics and a B.S. in Economics from MIT in 2011, and a Ph.D. in Economics from Harvard in 2016.