Mind Over Data: Introduction to Causal Inference
Abstract: Classical approaches to statistics and to machine learning are “cause blind,” in the sense that they analyze the data and answer queries about it without reference to the causal mechanism that produces the data. But the questions society asks of scientists are mostly causal questions: Will regular exercise prevent me from having a heart attack? Does smoking cause cancer? Without the tools to model causation explicitly, scientists struggled for years to answer such questions. In the case of smoking, their inability to frame an answer came at a high social cost.
There is no longer any need for scientists to observe a taboo on causal language. This lecture will provide a short conceptual introduction to the causal modeling framework developed by Judea Pearl, which relies on simple and intuitive dot-and-arrow diagrams to represent causal assumptions. We will give examples of the kinds of questions they can answer, and see how they resolve paradoxes that are difficult or impossible to unravel without a causal model.
Bio: Dana Mackenzie, a mathematician turned science writer, is the co-author, with Judea Pearl, of The Book of Why: The New Science of Cause and Effect, published by Basic Books in May 2018. His previous books include The Big Splat, or How Our Moon Came to Be (2003) and The Universe in Zero Words (2012). The former book was named an Editor’s Choice by Booklist, and the latter was a finalist for the 2017 Premio Asimov (Asimov Prize) in Italy. In addition, Mackenzie has written several volumes of What’s Happening in the Mathematical Sciences, an ongoing series published by the American Mathematical Society. He has been awarded the Chauvenet Prize for exposition by the Mathematical Association of America, and the Communication Award by the Joint Policy Board for Mathematics.
