Using explainability to understand ML models and reduce bias

We introduced the concept of explainability at a high level in the previous section. This section dives further into this topic, introducing tools that can be used to gain insights into how ML models are working at inference time.

Explainability techniques, methods, and tools

Let’s begin by exploring some popular techniques, methods, and tools that we can use for implementing explainability in ML, which we describe in the following subsections.

Performing data exploration

By now, it should hopefully be clear that understanding the data used to train our models is one of the first steps in explaining how the model makes decisions, and it is also one of the first lines of defense to identify and combat potential biases.

In the practical activities associated with this chapter, we explore the “Adult Census Income” dataset (https://archive.ics.uci.edu/dataset/2/adult), which is known to contain...