The Counterfactual Explanations Method
The EU General Data Protection Regulation (GDPR) states that an automatic process acting on personal data must be explained. Chapter 2, White Box XAI for AI Bias and Ethics, described the legal obligations that artificial intelligence (AI) faces. If a plaintiff requires an explanation for a decision that a machine learning (ML) application made, the editor of the AI system must justify its output.
In the previous chapters, we used a bottom-to-top method to understand an explainable AI (XAI) method. We displayed the source code. We then saw how the data was loaded, vectorized, preprocessed, and split into the training/test datasets. We described the mathematical representation of some of the ML and deep learning algorithms in the programs. We also defined the parameters of the ML models.
In Chapter 4, Microsoft Azure Machine Learning Model Interpretability with SHAP, we calculated the Shapley value of a feature to determine its...
