Learning Causal Graphical Models of Large-Scale Systems
Effective methods for inferring causal dependence from observational data have been developed within both computer science and quantitative social science. Methods in computer science have focused on the correspondence between casual graphical models and observed patterns of statistical association. Methods in social science have focused on templates for causal inference often called quasi-experimental designs, including instrumental variables, propensity scores matching, regression discontinuinty designs, and interrupted time-series designs. Recent work has begun to unify these methods using the framework of causal graphical models, but barriers remain because the formal framework of causal graphical models has been insufficiently expressive to represent some quasi-experimental designs. In this talk, I will introduce many of the known experimental and quasi-experimental designs in the language of directed graphical models, and I will present very recent work on defining additional designs in terms of classes of graphical models that can represent relational and temporal dependence. Finally, I will present two novel designs that have resulted from our work on causal inference in relational data.
Bio: David Jensen is Professor of Computer Science and Director of the Knowledge Discovery Laboratory at the University of Massachusetts Amherst. He received his doctorate from Washington University in St. Louis in 1992. From 1991 to 1995, he served as an analyst with the Office of Technology Assessment, an agency of the United States Congress. His research focuses on machine learning and causal inference in relational data sets, with applications to social network analysis, computational social science, fraud detection, and management of large technical systems. He has served on the Executive Committee of the ACM Special Interest Group on Knowledge Discovery and Data Mining and on the program committees of the International Conference on Machine Learning, the International Conference on Knowledge Discovery and Data Mining, and the Uncertainty in AI Conference. He was a member of the 2006-2007 Defense Science Study Group, and served for six years on DARPA's Information Science and Technology (ISAT) Group. He serves as Associate Director of the UMass Computational Social Science Institute. He won the 2011 Outstanding Teaching Award from the UMass College of Natural Science.