Causality, Counterfactuals and AI

Catalog description

The ability to reason about what might have been is one of the most central aspects of intelligence, and is a key part of what enables people to generalize from prior experience to inform their future decisions. This issue has captivated multiple communities and also is central to areas from healthcare to economics. In this course we will introduce the dominant approaches in machine learning and AI, with also reference to statistics and econometrics. Classes will combine lectures and discussions. Assignments will involve reasoning about the alternate frameworks and the questions they can address, using presented approaches to infer treatment effects in existing datasets, and essays arguing in favor of one of the particular frameworks for causal and counterfactual reasoning.