What Counterfactuals Can Be Tested

Abstract

Counterfactual statements, e.g., "my headache would be gone had I taken an aspirin" are central to scientific discourse, and are formally interpreted as statements derived from "alternative worlds". However, since they invoke hypothetical states of affairs, often incompatible with what is actually known or observed, testing counterfactuals is fraught with conceptual and practical difficulties. In this paper, we provide a complete characterization of "testable counterfactuals," namely, counterfactual statements whose probabilities can be inferred from physical experiments. We provide complete procedures for discerning whether a given counterfactual is testable and, if so, expressing its probability in terms of experimental data.

Appears in Proceedings of the Twenty-Third Conference on Uncertainty in Artificial Intelligence (UAI2007)

References in corpus (1)

• Probabilistic Evaluation of Sequential Plans from Causal Models with Hidden Variables

Cited by in corpus (10)

• On the Validity of Covariate Adjustment for Estimating Causal Effects
• Effects of Treatment on the Treated: Identification and Generalization
• Survey on Causal-based Machine Learning Fairness Notions
• Partial Counterfactual Identification from Observational and Experimental Data
• Nested Counterfactual Identification from Arbitrary Surrogate Experiments
• Identifying Counterfactual Queries with the R Package cfid
• Estimation of Counterfactual Interventions under Uncertainties
• Simulating counterfactuals
• Neural Networks for Learning Counterfactual G-Invariances from Single Environments
• Algorithmic Causal Effect Identification with causaleffect