[Paper Review] Applied Causal Inference Powered by ML and AI

Victor Chernozhukov, Christian Hansen+3 more|arXiv (Cornell University)|Mar 4, 2024

Advanced Data Processing Techniques36 citations

TL;DR

This paper introduces the fusion of machine learning and causal inference, connecting SEMs with DAGs/SCMs and discussing Double/Debiased Machine Learning for inference using modern predictive tools.

ABSTRACT

An introduction to the emerging fusion of machine learning and causal inference. The book presents ideas from classical structural equation models (SEMs) and their modern AI equivalent, directed acyclical graphs (DAGs) and structural causal models (SCMs), and covers Double/Debiased Machine Learning methods to do inference in such models using modern predictive tools.

Motivation & Objective

• Motivate the integration of classical causal inference models with modern AI frameworks.
• Explain how SEMs relate to DAGs and SCMs in the context of ML methods.
• Summarize Double/Debiased Machine Learning techniques for inference in causal models using predictive tools.

Proposed method

• Discuss classical structural equation models (SEMs) and their modern AI counterparts, directed acyclic graphs (DAGs) and structural causal models (SCMs).
• Present Double/Debiased Machine Learning methods for conducting inference in these models.
• Highlight the use of contemporary predictive tools within causal inference workflows.

Experimental results

Research questions

• RQ1 How can classical SEMs be integrated with DAGs/SCMs in the era of ML and AI?
• RQ2 What role do Double/Debiased Machine Learning methods play in causal inference with modern predictors?
• RQ3 How can predictive tools be leveraged for inference in causal models?
• RQ4 What are the practical implications of fusing ML and causal inference for econometric analysis?

Key findings

• The paper outlines a fusion of machine learning with causal inference terminology and frameworks.
• It connects SEMs with their AI equivalents, DAGs and SCMs.
• It discusses the use of Double/Debiased Machine Learning for inference in causal models using modern predictive tools.