Abstract:This paper serves as a starting point for machine learning researchers, engineers and students who are interested in but not yet familiar with causal inference. We start by laying out an important set of assumptions that are collectively needed for causal identification, such as exchangeability, positivity, consistency and the absence of interference. From these assumptions, we build out a set of important causal inference techniques, which we do so by categorizing them into two buckets; active and passive approaches. We describe and discuss randomized controlled trials and bandit-based approaches from the active category. We then describe classical approaches, such as matching and inverse probability weighting, in the passive category, followed by more recent deep learning based algorithms. By finishing the paper with some of the missing aspects of causal inference from this paper, such as collider biases, we expect this paper to provide readers with a diverse set of starting points for further reading and research in causal inference and discovery.

What problem does this paper attempt to address?

The problem that this paper attempts to solve is how to introduce the concepts and methods of causal inference (CI) in the field of machine learning, especially for researchers, engineers and students who are not yet familiar with causal inference. Specifically, the paper focuses on the following aspects: 1. **Basic assumptions for causal identification**: The paper first expounds several important assumptions required for causal identification, such as exchangeability, positivity, consistency and absence of interference. These assumptions are the basis for understanding causal relationships. 2. **Classification of causal inference techniques**: Based on the above assumptions, the paper divides important causal inference techniques into two categories: active and passive methods. Active methods include Randomized Controlled Trials (RCT) and bandit - based approaches; passive methods include classic methods such as matching, inverse probability weighting, and recent deep - learning - based algorithms. 3. **Difference between intervention and condition**: The paper emphasizes the difference between observation and intervention. Observation is to passively record the occurrence of events, while intervention is to actively change the state of a variable and observe its impact. This distinction is crucial for understanding causal relationships. 4. **Potential outcome framework and causal graph**: The paper combines Rubin's potential outcome framework and Pearl's do - calculus framework to provide a comprehensive basis for causal inference. At the same time, the paper introduces the concept of causal graph, which is a graphical method for representing causal relationships between variables. 5. **Data collection and experimental design**: The paper discusses how to meet the basic assumptions of causal inference through active or passive data collection methods. For example, RCT is an active data collection method that can effectively avoid the influence of confounding factors. 6. **Challenges in practical applications**: The paper also explores the challenges that may be encountered in practical applications, such as how to deal with high - dimensional data and how to generalize when the data does not meet the positivity assumption. In short, this paper aims to provide readers with a comprehensive introductory guide to causal inference, helping them understand and apply the methods and techniques of causal inference.

Active and Passive Causal Inference Learning

Causal Inference Meets Machine Learning

Causal Inference and Counterfactual Prediction in Machine Learning for Actionable Healthcare

From Dependence to Causation

Methods and tools for causal discovery and causal inference

Deep Causal Learning: Representation, Discovery and Inference

A Survey on Causal Inference

Causal Inference and Related Statistical Methods

Statistical Approaches for Causal Inference

Causal Machine Learning: A Survey and Open Problems

Towards Understanding How Machines Can Learn Causal Overhypotheses

Passive learning of active causal strategies in agents and language models

Causal Inference: A Missing Data Perspective

Instrumental Variables in Causal Inference and Machine Learning: A Survey

Causal discovery and inference: concepts and recent methodological advances

Applied Causal Inference Powered by ML and AI

Causal learner: A toolbox for causal structure and Markov blanket learning

Introducing CausalBench: A Flexible Benchmark Framework for Causal Analysis and Machine Learning

Evaluation Methods and Measures for Causal Learning Algorithms

Causal inference in AI education: A primer

Causal inference and data fusion in econometrics