xai_evals : A Framework for Evaluating Post-Hoc Local Explanation Methods

The increasing complexity of machine learning (ML) and deep learning (DL) models has led to their widespread adoption in numerous real-world applications. However, as these models become more powerful, they also become less interpretable. In particular, deep neural networks (DNNs), which have achieved state-of-the-art performance in tasks such as image recognition, natural language processing, and autonomous driving, are often viewed as "black box" models due to their complexity and lack of transparency. Interpretability is essential, particularly in high-stakes fields where the consequences of incorrect or non-explainable decisions can be profound. In domains such as healthcare, finance, and law, it is not only crucial that AI systems make accurate predictions but also that these predictions can be understood and justified by human stakeholders. For example, in healthcare, understanding why a model predicts a certain diagnosis can be as important as the prediction itself, influencing clinical decisions and patient outcomes.