Chapter 3: Data-Centric Approaches
In the Defining explanation methods and approaches section of Chapter 1, Foundational Concepts of Explainability Techniques, when we looked at the various dimensions of explainability, we discussed how data is one of the important dimensions. In fact, all machine learning (ML) algorithms depend on the underlying data being used.
In the previous chapter, we discussed various model explainability methods. Most of the methods discussed in Chapter 2, Model Explainability Methods, are model-centric. The concepts and ideas discussed were focused on making black-box models interpretable. But recently, the ML and AI communities have realized the core importance of data for any analysis or modeling purposes. So, more and more AI researchers are exploring new ideas and concepts around data-centric AI.
Since data plays a vital role in the model-building and prediction process, it is even more important for us to explain the functioning of any ML and AI...