GRANITE: A Generalized Regional Framework for Identifying Agreement in Feature-Based Explanations

2026-01-30

AI Summary

This work addresses the inconsistency and even contradictions among existing feature attribution methods, which often arise from differing treatments of feature interactions and dependencies. To resolve this, we propose GRANITE, a framework that recursively partitions the feature space to automatically identify regions where both interaction effects and distributional shifts are minimized, thereby enabling consistent alignment of multiple explanation methods within each region. GRANITE unifies and extends current regionalized explanation strategies to the level of feature groups, substantially improving agreement across diverse attribution techniques. Experimental results on multiple real-world datasets demonstrate that GRANITE effectively enhances the reliability and interpretability of feature attributions, offering a practical tool for trustworthy model explanations in real-world applications.

Technology Category

• Computer Vision: Interpretability, Explainability, and Transparency
• Machine Learning: Feature Construction/Reformulation
• Constraint Satisfaction and Optimization: Satisfiability Modulo Theories

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
• Web Mining and Content Analysis: Robustness and generalizability of Web mining methods

Abstract

Feature-based explanation methods aim to quantify how features influence the model's behavior, either locally or globally, but different methods often disagree, producing conflicting explanations. This disagreement arises primarily from two sources: how feature interactions are handled and how feature dependencies are incorporated. We propose GRANITE, a generalized regional explanation framework that partitions the feature space into regions where interaction and distribution influences are minimized. This approach aligns different explanation methods, yielding more consistent and interpretable explanations. GRANITE unifies existing regional approaches, extends them to feature groups, and introduces a recursive partitioning algorithm to estimate such regions. We demonstrate its effectiveness on real-world datasets, providing a practical tool for consistent and interpretable feature explanations.

Problem

• feature-based explanations
• explanation disagreement
• feature interactions
• feature dependencies
• interpretability

Innovation

• feature-based explanation
• regional framework
• feature interaction
• recursive partitioning
• explanation consistency