A Sound and Complete Algorithm for Learning Causal Models from Relational Data

TL;DR

This paper introduces RCD, a sound and complete algorithm for learning causal relational models from relational data, leveraging lifted reasoning for structure orientation, with proven theoretical guarantees and empirical validation.

Contribution

The paper develops RCD, the first complete algorithm for relational causal model learning, extending causal discovery to more expressive relational data.

Findings

RCD is proven to be sound and complete.

Empirical results demonstrate RCD's effectiveness.

RCD advances causal discovery in relational domains.

Abstract

The PC algorithm learns maximally oriented causal Bayesian networks. However, there is no equivalent complete algorithm for learning the structure of relational models, a more expressive generalization of Bayesian networks. Recent developments in the theory and representation of relational models support lifted reasoning about conditional independence. This enables a powerful constraint for orienting bivariate dependencies and forms the basis of a new algorithm for learning structure. We present the relational causal discovery (RCD) algorithm that learns causal relational models. We prove that RCD is sound and complete, and we present empirical results that demonstrate effectiveness.