Towards a Unified Multidimensional Explainability Metric: Evaluating Trustworthiness in AI Models

Georgios Makridis, Georgios Fatouros, Athanasios Kiourtis +4

The paper proposes a framework that evaluates explainability methods like LIME and SHAP across models and datasets using fidelity, simplicity, and stability, and builds a knowledge…