1 Why causal AI

This chapter covers

Defining causal AI and its benefits
Incorporating causality into machine learning models
A simple example of applying causality to a machine learning model

Subscription streaming platforms like Netflix are always looking for ways to optimize various indicators of performance. One of these is their churn rate, meaning the rate at which they lose subscribers. Imagine that you are a machine learning engineer or data scientist at Netflix tasked with finding ways of reducing churn. What are the types of causal questions (questions that require causal thinking) you might ask with respect to this task?

Causal discovery —Given detailed data on who churned and who did not, can you analyze that data to find causes of the churn? Causal discovery investigates what causes what.
Estimating average treatment effects (ATEs)—Suppose the algorithm that recommends content to the user is a cause of the churn; a better choice of algorithm might reduce churn, but by how much? The task of quantifying how much, on average, a cause drives an effect is the ATE estimation. For example, some users could be exposed to a new version of the algorithm, and you could measure how much this affects churn, relative to the baseline algorithm.

1.1 What is causal AI?

1.2 How this book approaches causal inference

1.2.1 Emphasis on AI

1.2.2 Focus on tech, retail, and business

1.2.3 Parallel world counterfactuals and other queries beyond causal effects

1.2.4 An assumption of commodification of inference

1.2.5 Breaking down theory with code

1.3 Causality’s role in modern AI workflows

1.3.1 Better data science

1.3.2 Better attribution, credit assignment, and root cause analysis

1.3.3 More robust, decomposable, and explainable models

1.3.4 Fairer AI

1.4 How causality is driving the next AI wave

1.4.1 Causal representation learning