Laplace’s Causal Demon

Bayesian inference is a general tool for decision making in an uncertain world.  The modern formulation due to Ramsey, de Finetti and Savage provides an axiomization for decision making under uncertainty.  Bayesian inference is characterized by rigid principles but flexible assumptions and it is at the heart of modern artificial intelligence.  Causal inference is a largely parallel theoretical development which focuses on identification of causal effect, prediction and ultimately optimization of the effect of interventions.  Curiously many of the most useful causal inference techniques are not easy to motivate using Bayesian principles.  For example the widely used Horvitz-Thompson estimator can be used to estimate causal effects and has excellent frequentist properties but appears to violate the conditionality principle.  Similarly, Pearl’s do-calculus augments probability theory with additional rules to adapt to causal applications. This webinar series will be an exploration of the intersection of Bayesian inference and causal inference.  Our speakers will help us understand how we can use these two frameworks in order to solve applied problems, and will consider if these different frameworks are in conflict or are complimentary.

The audience is machine learning practitioners and statisticians from academia and industry.

To stay informed follow us on twitter, or we have a Google Group for general announcements and discussions related to the seminar series. Join the group here.