Open Source XAI Libraries: Trust Analysis of SHAP, LIME, DiCE, and Alibi

Key Points

• This research aims to analyze the trustworthiness of popular open-source explainable artificial intelligence (XAI) libraries.
• Evaluated trust metrics across SHAP, LIME, DiCE, and Alibi libraries.
• Conducted comparative analysis of interpretability and reliability.
• Utilized case studies and benchmarks for assessment.
• Identified varying levels of trust effectiveness across different libraries.
• SHAP demonstrated superior trust metrics compared to LIME and Alibi.
• DiCE showed potential for enhanced trust in specific contexts.

Abstract