Deep Learning in Banking

Chapter 8Fairness, Accountability, Explainability, and Causality

Summary

After reading this chapter, and going through the exercises, you will be able to:

• Understand the concepts of fairness, accountability, transparency, and explainability and how they appear in banking.
• Understand causality and its relevance in banking applications.
• Measure model fairness and use SHAP to explain text-based models.

This book on deep learning in banking shows how to leverage artificial intelligence (AI) for next-generation financial services. It has shown how to incorporate various sources of data into complex deep learning architectures and how that can lead to impressive performance gains when predicting (i.e., credit scores and delinquency). Financial institutions have started unlocking the potential of alternative data sources and sophisticated models to make sense of complexity, and thus improve credit risk management and other operations, as the benefits within the banking sector are obvious.

However, there are substantial risks associated with using these techniques and machine learning in general, as they can enhance structural inequalities, and thus greatly limit access to credit and other financial opportunities for people who are historically or culturally underrepresented. An even greater problem with fairness, in particular, is that it has not been properly defined, which means that it is challenging to truly account for such inequalities in a robust and systematic way, ...