Prediction in social science

Feb. 18, 2015 from 3:30 to 4:20 p.m. — Physics/Astronomy Auditorium, room A102

Professor, Harvard University


I have spent the last few years working with computer scientists (notably Jon Kleinberg) to integrate machine learning in to social science applications broadly construed and behavioral economics specifically. In this talk I will describe the high points (and low points) of this work. Along the way we have struggled with inherent tensions between predictive models and the broader goals of social science. The two primary tensions are: (1) How can we apply these tools for decision making when they (effectively) provide correlation but not causation?; and (2) How can we integrate them with theory testing and building? Solving these kinds of questions raise interesting challenges for the field. I will sketch potential solutions and argue that these problems look solvable. I will argue there is great research synergy between machine learning and social and behavioral sciences. These tools have the potential to revolutionize our understanding of human behavior and to open up fruitful areas of machine learning research. Applications to criminal justice, finance and behavioral economics will illustrate these points.


Sendhil Mullainathan is a Professor of Economics at Harvard University. His real passion is behavioral economics. His work runs a wide gamut: the impact of poverty on mental bandwidth; whether CEO pay is excessive; using fictitious resumes to measure discrimination; showing that higher cigarette taxes makes smokers happier; modeling how competition affects media bias; and a model of coarse thinking. His latest research focuses on using machine learning and data mining techniques to better understand human behavior.

He enjoys writing, having recently co-authored Scarcity: Why Having too Little Means so Much and writes regularly for the New York Times.

He helped co-found a non-profit to apply behavioral science (ideas42), co-founded a center to promote the use of randomized control trials in development (the Abdul Latif Jameel Poverty Action Lab), serves on the board of the MacArthur Foundation, and has worked in government in various roles, including most recently as Assistant Director of Research at the Consumer Financial Protection Bureau.

He is a recipient of the MacArthur “genius” Award, has been designated a “Young Global Leader” by the World Economic Forum, labeled a “Top 100 Thinker” by Foreign Policy Magazine, and named to the “Smart List: 50 people who will change the world” by Wired Magazine (UK). His hobbies include basketball, board games, googling and fixing-up classic espresso machines.