What Counterfactuals Can Be Tested

By Ilya Shpitser, Judea Pearl. 2007

127 citations recorded by the source corpus.

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.

Subfields: Artificial Intelligence, General Decision Sciences

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Artificial Intelligence

In General Decision Sciences

Citations in the corpus

Improving the accuracy of medical diagnosis with causal machine learning (2020, methodology): “Note that counterfactuals cannot in general be identi ﬁ ed from data alone, and require modelling assumptions such as knowledge of the underlying structural equations 58,64 .”
Counterfactual Normalization: Proactively Addressing Dataset Shift and Improving Reliability Using Causal Mechanisms (2018, background): “Other graphical representations such as twin networks (Pearl, 2009) and counterfactual graphs (Shpitser and Pearl, 2007) simultaneously represent factual and coun-terfactual outcomes, rather than the intermediate counter-factuals exploited in this work.”
Causal inference in statistics: An overview (2009, background): “…and sufficient condition for identification (Shpitser and Pearl, 2006), and (2) extending 10Before applying this criterion, one may delete from the causal graph all nodes that are not ancestors of Y . the condition from causal effects to any counterfactual expression (Shpitser and Pearl, 2007).”
Effects of Treatment on the Treated: Identification and Generalization (2009, methodology): “The resulting graph containing these copies is known as a counterfactual graph [10].”
PC-Fairness: A Unified Framework for Measuring Causality-based Fairness (2019, background): “When O 6= ∅ and π = Π, the reason for unidentifiability can be the existence of the “w graph” (see Figure 4) [11].”
Single World Intervention Graphs ( SWIGs ) : A Unification of the Counterfactual and Graphical Approaches to Causality (2013, methodology): “…graphs and counterfactuals with minimal assumptions As just noted, our approach differs from previously proposed attempts to link graphs and counterfactuals such as the ‘twin network’ approach of Balke and Pearl (1994); Pearl (2000) and the ‘counterfactual graph’ of Shpitser and Pearl (2007, 2008).”
Complete Identification Methods for the Causal Hierarchy (2008, background): “These variables are functionally dependent on some subset of U [ V .”
The Causal Foundations of Structural Equation Modeling (2012, background): “For example, the absence of variable Z from the arguments of fY conveys the empirical claim that variations in Z will leave Y unchanged, as long as variables UY and X remain constant.”