Causality: Counterfactuals | Part A

Tutorial on causal inference, covering the basics of counterfactual thinking. Topics: causal mechanisms and why we need them for counterfactual reasoning; types of models needed for counterfactual reasoning: functional Bayesian networks and structural causal models; probabilities of causation; twin-network technique; why interventional reasoning is not as refined as counterfactual reasoning; syntax and semantics of counterfactual queries; open-ended counterfactual queries; and identifiability of counterfactual queries (point estimates and bounds) based on observational and experimental data. Includes discussion of prototypical counterfactual queries such as probability of necessity (PN), probability of sufficiency (PS) and probability of necessity and sufficiency (PNS). 00:00 The Causal Hierarchy 01:33 Examples of Counterfactual Queries 04:31 Agenda: Idealized and Practical Settings For Computing Counterfactuals 07:00 The Information Hierarchy 08:31 Notation For Counterfactuals 09:16 Functional Dependencies: The Information Needed for Counterfactual Reasoning 13:48 Why We Need Functional Dependencies (Causal Mechanisms) 16:59 From Bayesian Networks To Functional Bayesian Networks 18:58 Functional Bayesian Networks (Structural Causal Models) 20:16 Exogeneity: Exchanging Actions and Observations 23:02 Probability Of Necessity (PN) 29:57 Probability Of Sufficiency (PS) 32:14 Twin-Network Technique For Computing Counterfactual Queries 34:12 Probability of Necessity and Sufficiency (PNS) 37:23 Key Insights About Counterfactual Reasoning 39:44 Monotonicity: No Contrarians 41:49 Why Is Counterfactual Reasoning More Powerful Than Interventional Reasoning 45:03 Local vs Global Counterfactuals --- Slides available at: http://web.cs.ucla.edu/~darwiche/caus... --- On Pearl's Causal Hierarchy and the Foundations of Causal Inference: https://causalai.net/r60.pdf --- Causal Inference Using Tractable Circuits: https://arxiv.org/pdf/2202.02891.pdf