IDSS Distinguished Seminar – Peter Spirtes (Carnegie Mellon University)
This talk gives a broad overview of central concepts and principles involved in automated causal inference and emerging approaches to causal discovery from observational data or combinations of observational and experimental data. I will describe several approaches to automated causal inference, and discuss the plausibility and usefulness of various alternative assumptions underlying their validity. I will then focus on the major limitations of current automated causal inference algorithms, and how those limitations might be addressed.
Professor Spirtes is a co-author of “Discovering Causal Structure”, “Causation, Prediction, and Search”, the TETRAD program, and numerous articles on automated discovery of causal relations from combinations of background knowledge, and non-experimental and experimental data. One goal of his research is to specify and prove under what conditions it is possible to reliably infer causal relationships from background knowledge and statistical data not obtained under fully controlled conditions. A second goal is to develop, analyze, implement, test and apply practical, provably correct computer programs for inferring causal structure under conditions where this is possible. The results of this research are available in the TETRAD computer program.