American Statistical Association
New York City
Metropolitan Area Chapter

Levin Lecture Series: Fall 2019 Colloquium Seminars
Department of Biostatistics
Mailman School of Public Health
Columbia University



MACHINE LEARNING IS NATURAL EXPERIMENT

by

Dr. Yusuke Narita
Assistant Professor, Economics
Yale University

Host: Ying Wei


Abstract

Machine learning algorithms produce a growing portion of decisions and recommendations. Such algorithmic decisions are natural experiments (conditionally quasi-randomly assigned instrumental variables) since the algorithms make decisions based only on observable variables. We formalize this observation and characterize the sources of identification of causal effects for a class of supervised learning, reinforcement learning, and bandit algorithms. Data from almost every machine learning algorithm is shown to identify some causal effect. This identification result translates into easily-implemented instrumental variable estimators of causal effects that are free from the curse of dimensionality. We provide empirical applications to the tech industry and public policy.


Date: Thursday, November 14, 2019
Time: 11:30 A.M. - 12:30 P.M.
Location: Mailman School of Public Health
Department of Biostatistics
722 West 168th Street
AR Building
8th Floor Auditorium
New York, New York

Home Page | Chapter News | Chapter Officers | Chapter Events
Other Metro Area Events | ASA National Home Page | Links To Other Websites
NYC ASA Chapter Constitution | NYC ASA Chapter By-Laws

Page last modified on November 6, 2019
Copyright © 1998-2019 by New York City Metropolitan Area Chapter of the ASA
Designed and maintained by Cynthia Scherer
Send questions or comments to nycasa@nycasa.org