A Logical Characterization of Constraint-Based Causal Discovery

TL;DR

This paper introduces a logical inference-based method for constraint-based causal discovery that is sound, complete, and effective even with latent variables and selection bias, simplifying causal relation identification.

Contribution

It presents a new logical framework for causal discovery that does not depend on detailed graph structures, enabling more robust and scalable causal inference.

Findings

Identifies all invariant features of the PAG.

Shows every causal relation has one of two fundamental forms.

Applicable to large models with latent variables.

Abstract

We present a novel approach to constraint-based causal discovery, that takes the form of straightforward logical inference, applied to a list of simple, logical statements about causal relations that are derived directly from observed (in)dependencies. It is both sound and complete, in the sense that all invariant features of the corresponding partial ancestral graph (PAG) are identified, even in the presence of latent variables and selection bias. The approach shows that every identifiable causal relation corresponds to one of just two fundamental forms. More importantly, as the basic building blocks of the method do not rely on the detailed (graphical) structure of the corresponding PAG, it opens up a range of new opportunities, including more robust inference, detailed accountability, and application to large models.