We formalize the concept of verification in causal graphical models: determining whether a given observational formula identifies a target interventional distribution. This introduces a problem complementary to identification, focusing not on whether any identifying formula exists, but whether the given formula is indeed identifying. We demonstrate that even sound and complete solutions to identification do not suffice for verification. We propose a falsifier as a practical first step, proving it induces an almost-surely correct verifier for regular exponential-family models, and use this verifier to develop the gateway test, which identifies all sets admissible for use in a front-door formula.
Blogger's Review: This study offers a fresh perspective in the realm of causal inference by emphasizing the distinction between verification and identification, and advances theoretical development through the construction of effective verifiers, potentially playing a crucial role in the analysis of complex causal models.