An increasing number of decisions are guided by machine learning algorithms. But when consequential decisions are encoded in algorithms, individuals may strategically alter their behavior to achieve desired outcomes. This paper develops an empirical approach to adjust decision algorithms to anticipate manipulation. By explicitly modeling incentives to manipulate, our approach produces decision rules that are stable under manipulation, even when the rules are fully transparent. We stress test this approach through a large field experiment with smartphone users in Kenya, designed to mimic the “digital loan” services that have recently proliferated in East Africa. When implemented, decision rules estimated with this strategy-robust approach outperform those based on standard machine learning approaches.

Bio:  Joshua Blumenstock is a Chancellor’s Associate Professor at the U.C. Berkeley School of Information and the Goldman School of Public Policy. He is the Co-director of the Global Policy Lab and the Center for Effective Global Action. Blumenstock does research at the intersection of machine learning and empirical economics, with a focus on how novel data and technology can better address the needs of poor and vulnerable people around the world. He has a Ph.D. in Information Science and a M.A. in Economics from U.C. Berkeley, and Bachelor’s degrees in Computer Science and Physics from Wesleyan University. He is a recipient of awards including the NSF CAREER award, the Intel Faculty Early Career Honor, and the U.C. Berkeley Chancellor’s Award for Public Service. His work has appeared in general interest journals including Science, Nature, and Proceedings of the National Academy of Sciences,as well as top economics journals (e.g., the American Economic Review) and computer science conferences (e.g., ICML, KDD, AAAI, WWW, CHI).