Informative Perturbation Selection for Uncertainty-Aware Post-hoc Explanations

2026-03-16

AI Summary

This work addresses the challenge of efficiently constructing local neighborhoods to generate reliable, reproducible explanations with uncertainty estimates in post-hoc interpretability settings where internal model information is unavailable. The authors propose EAGLE, a novel framework that formulates perturbation selection as an information-theoretic active learning problem. By adaptively sampling perturbations that maximize expected information gain, EAGLE learns a linear surrogate model and outputs both feature importance scores and their associated confidence intervals. Theoretical analysis demonstrates that the cumulative information gain grows logarithmically, with sample complexity exhibiting a linear-logarithmic relationship. Empirical evaluations on tabular and image data show that EAGLE significantly outperforms state-of-the-art methods—including Tilia, US-LIME, GLIME, and BayesLIME—in terms of explanation reproducibility, neighborhood stability, and perturbation quality.

Technology Category

• Machine Learning: Active Learning
• Reasoning under Uncertainty: Uncertainty Representations
• Humans and AI: Explainable AI (XAI) for Human Understanding

Application Category

• User Modeling, Personalization and Recommendation: Explainable and interpretable methods for personalization
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Semantics and Knowledge: Data modeling to support human-machine intelligence, including LLMs agents, intelligent system behavior, explanations, and user-friendly interactions

Abstract

Trust and ethical concerns due to the widespread deployment of opaque machine learning (ML) models motivating the need for reliable model explanations. Post-hoc model-agnostic explanation methods addresses this challenge by learning a surrogate model that approximates the behavior of the deployed black-box ML model in the locality of a sample of interest. In post-hoc scenarios, neither the underlying model parameters nor the training are available, and hence, this local neighborhood must be constructed by generating perturbed inputs in the neighborhood of the sample of interest, and its corresponding model predictions. We propose \emph{Expected Active Gain for Local Explanations} (\texttt{EAGLE}), a post-hoc model-agnostic explanation framework that formulates perturbation selection as an information-theoretic active learning problem. By adaptively sampling perturbations that maximize the expected information gain, \texttt{EAGLE} efficiently learns a linear surrogate explainable model while producing feature importance scores along with the uncertainty/confidence estimates. Theoretically, we establish that cumulative information gain scales as $\mathcal{O}(d \log t)$, where $d$ is the feature dimension and $t$ represents the number of samples, and that the sample complexity grows linearly with $d$ and logarithmically with the confidence parameter $1/δ$. Empirical results on tabular and image datasets corroborate our theoretical findings and demonstrate that \texttt{EAGLE} improves explanation reproducibility across runs, achieves higher neighborhood stability, and improves perturbation sample quality as compared to state-of-the-art baselines such as Tilia, US-LIME, GLIME and BayesLIME.

Problem

• post-hoc explanation
• uncertainty-aware
• perturbation selection
• model-agnostic
• black-box model

Innovation

• post-hoc explanation
• active learning
• information gain
• uncertainty quantification
• model-agnostic interpretability