ConceptSMILE: Auditing the Trustworthiness of Concept-Based Explainable AI

2026-07-10

AI Summary

Current concept-level explainable AI methods lack an automated mechanism for credibility assessment. This work proposes ConceptSMILE—a model-agnostic perturbation auditing framework that extends input perturbation logic from feature attribution to human-interpretable concept explanations. By applying locally weighted perturbations, measuring shifts in concept responses, and fitting an XGBoost surrogate model, ConceptSMILE establishes an independent credibility evaluation layer. Experiments on fundus images demonstrate that MedSAM-derived concepts achieve superior performance in spatial attribution and surrogate fidelity (R² = 0.8503), whereas vision-language model (VLM)-derived concepts exhibit greater faithfulness to vascular structures and enhanced robustness under specific artifacts.

Abstract

Concept-based explainable artificial intelligence (AI) can make model reasoning more human-understandable, but concept-level outputs are not automatically trustworthy. We introduce ConceptSMILE, a model-agnostic perturbation-based auditing framework for evaluating the reliability of concept-based explanations. Rather than replacing SMILE, ConceptSMILE extends its perturbation-based logic from feature- or region-level attribution to the auditing of human-understandable concept explanations. The framework perturbs input regions, measures concept-response shifts, applies locality weighting, and fits an XGBoost surrogate to approximate local concept behaviour. Reliability is assessed through attribution accuracy, surrogate fidelity, faithfulness, stability, and consistency. We evaluate ConceptSMILE on retinal fundus images by comparing MedSAM-derived visual concepts with VLM-based semantic concepts. Results show that reliability varies across concepts and pathways: MedSAM achieves stronger spatial attribution and the highest surrogate fidelity ($R^2 = 0.8503$, $R_w^2 = 0.8465$), while the VLM pathway shows stronger vessel faithfulness and stronger stability under selected artefact conditions. ConceptSMILE provides an independent audit layer for evaluating the trustworthiness of concept-based XAI.

Problem

• concept-based explainable AI
• trustworthiness
• reliability
• concept explanations
• auditing

Innovation

• concept-based XAI
• perturbation-based auditing
• surrogate fidelity
• model-agnostic framework
• trustworthiness evaluation