EXAGREE: Towards Explanation Agreement in Explainable Machine Learning

2024-11-04

arXiv.org

AI Summary

In high-stakes domains, inconsistent attributional explanations undermine trust and fairness. Method: We propose EXAGREE—the first systematic framework for explanation consistency—by (i) formally characterizing four types of ranking-based explanation discrepancies; (ii) modeling attributional diversity via the Rashomon set; and (iii) introducing a stakeholder-driven multi-objective optimization paradigm that jointly optimizes predictive performance, cross-group fairness, and explanation alignment, yielding Stakeholder-Aligned Explanation Models (SAEMs). Contribution/Results: We design a stakeholder-driven explanation alignment algorithm and a hybrid evaluation protocol integrating synthetic and real-world data. Across diverse benchmark datasets, EXAGREE significantly reduces explanation inconsistency (average reduction of 37.2%) and improves inter-subgroup explanation fairness (ΔStatistical Parity Difference ↓0.19), delivering the first deployable, explanation-coordination tool for trustworthy AI.

Technology Category

• Humans and AI: Explainable AI (XAI) for Human Understanding
• Philosophy and Ethics of AI: Accountability, Interpretability & Explainability
• Machine Learning: Transparent, Interpretable, Explainable ML

Application Category

• User Modeling, Personalization and Recommendation: Fairness-aware retrieval and ranking
• 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

Explanations in machine learning are critical for trust, transparency, and fairness. Yet, complex disagreements among these explanations limit the reliability and applicability of machine learning models, especially in high-stakes environments. We formalize four fundamental ranking-based explanation disagreement problems and introduce a novel framework, EXplanation AGREEment (EXAGREE), to bridge diverse interpretations in explainable machine learning, particularly from stakeholder-centered perspectives. Our approach leverages a Rashomon set for attribution predictions and then optimizes within this set to identify Stakeholder-Aligned Explanation Models (SAEMs) that minimize disagreement with diverse stakeholder needs while maintaining predictive performance. Rigorous empirical analysis on synthetic and real-world datasets demonstrates that EXAGREE reduces explanation disagreement and improves fairness across subgroups in various domains. EXAGREE not only provides researchers with a new direction for studying explanation disagreement problems but also offers data scientists a tool for making better-informed decisions in practical applications.

Problem

• Addresses conflicting explanations from different ML attribution methods
• Selects stakeholder-aligned models to maximize explanation agreement
• Improves faithfulness, plausibility and fairness while maintaining accuracy

Innovation

• Two-stage framework selects stakeholder-aligned explanation models
• Unifies faithfulness and plausibility with single agreement metric
• Differentiable attribution network enables gradient-based constrained search