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

arXiv cs.AI · 6 hours ago

Researchers developed a framework that evaluates explainability methods like LIME and SHAP across multiple datasets and machine learning models by measuring fidelity, simplicity, and stability. The framework was tested on three open-source datasets and creates a knowledge base that can estimate explainability scores for new, previously unseen datasets and models. The unified metric enables more systematic comparison of different explainability methods to support development of more transparent AI systems.