Toward Causal Representation Learning

TLDR

Machine learning and graphical causality developed separately, but cross‑pollination is growing as each field benefits from the other's advances, and a key challenge is causal representation learning—discovering high‑level causal variables from low‑level observations. The article reviews causal inference concepts and examines their relevance to machine learning challenges such as transfer and generalization, and outlines implications and key research directions at the intersection of the two fields. The authors conduct a review of causal inference concepts and map them onto machine learning open problems like transfer and generalization. The review identifies implications of causality for machine learning and proposes key research areas at their intersection.

Abstract

The two fields of machine learning and graphical causality arose and are developed separately. However, there is, now, cross-pollination and increasing interest in both fields to benefit from the advances of the other. In this article, we review fundamental concepts of causal inference and relate them to crucial open problems of machine learning, including transfer and generalization, thereby assaying how causality can contribute to modern machine learning research. This also applies in the opposite direction: we note that most work in causality starts from the premise that the causal variables are given. A central problem for AI and causality is, thus, causal representation learning, that is, the discovery of high-level causal variables from low-level observations. Finally, we delineate some implications of causality for machine learning and propose key research areas at the intersection of both communities.