Summary

In this chapter, you have learned about fairness in machine learning in different aspects. First, we discussed legal definitions of fairness and quantitative ways to measure these definitions. We then discussed technical methods to train models to meet fairness criteria. We also discussed causal models. We learned about SHAP as a powerful tool to interpret models and find unfairness in a model. Finally, we learned how fairness is a complex systems issue and how lessons from complex systems management can be applied to make models fair.

There is no guarantee that following all the steps outlined here will make your model fair, but these tools vastly increase your chances of creating a fair model. Remember that models in finance operate in high-stakes environments and need to meet many regulatory demands. If you fail to do so, damage could be severe.

In the next, and final, chapter of this book, we will be looking at probabilistic programming and Bayesian inference.