eXplainable AI ! … ?

In this talk I’ll provide an overview of my work in eXplainable AI (XAI), examining its state-of-the-art techniques, current trends, and limitations. I'll begin by introducing key XAI methods and explore the growing demand for fairness, accountability, and human-in-the-loop systems, as well as the challenges of balancing model accuracy with explainability. Despite significant progress, XAI faces limitations, including the trade-off between model complexity and interpretability, and the subjective nature of explanations.  I’ll also discuss the importance of Human-Centered AI, emphasizing that explanations must be understandable to people, and how insights from the social sciences can inform better explanation design.  And  finally, I will introduce Evaluative AI, a paradigm shift from the current model of XAI . This concept represents a step toward creating more accountable, robust, and transparent AI systems, ensuring that explanations not only make sense but also align with human values and decision-making needs.