Having examined advanced methods for identifying and estimating causal effects, we now address the practical challenge of embedding these methods within functioning machine learning systems. Standard ML workflows often prioritize predictive accuracy, sometimes overlooking the causal mechanisms that drive outcomes. This can limit a system's ability to support effective decision-making or predict the results of interventions.

This chapter provides techniques for incorporating causal thinking throughout the machine learning lifecycle. You will learn about:

• Applying causal knowledge to improve feature engineering and selection.
• Developing ML models that account for causal structure or objectives.
• Evaluating model performance using causal criteria, such as intervention effects or fairness.
• Using causal inference to refine the design and analysis of A/B tests.
• Monitoring deployed systems for shifts in underlying causal relationships.
• Architecting and maintaining causal inference components within MLOps frameworks.

The objective is to provide practical strategies for building ML systems that are not only predictive but also informed by causal analysis, leading to more dependable interventions and better system understanding.