Robust Counterfactual Explanations under Model Multiplicity Using Multi-Objective Optimization

2025-01-10

AI Summary

To address the inconsistency and unreliability of counterfactual explanations (CEs) across diverse machine learning models—stemming from model multiplicity—this paper proposes the first Pareto-improvement-based multi-objective optimization framework for CE generation. The framework jointly optimizes four objectives: cross-model consistency, action feasibility, low perturbation (L1) distance, and high predicted class confidence. Methodologically, it integrates NSGA-II for multi-objective optimization, model-agnostic CE generation, and novel robustness evaluation metrics. Its key innovation lies in introducing Pareto improvement into CE synthesis, thereby fundamentally mitigating model dependency and unifying explanation robustness with plausibility. Experiments on synthetic and real-world datasets demonstrate a 37.2% improvement in cross-model CE consistency while maintaining low L1 distance, high feasibility, and high prediction confidence.

Technology Category

• Machine Learning: Calibration & Uncertainty Quantification
• Constraint Satisfaction and Optimization: Satisfiability Modulo Theories
• Humans and AI: Explainable AI (XAI) for Human Understanding

Application Category

• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• User Modeling, Personalization and Recommendation: Attacks and countermeasures in recommendation systems
• Responsible Web: Machine-in-the-loop, human agency and autonomy

Abstract

In recent years, explainability in machine learning has gained importance. In this context, counterfactual explanation (CE), which is an explanation method that uses examples, has attracted attention. However, it has been pointed out that CE is not robust when there are multiple machine-learning models. These problems are important when using machine learning to make safe decisions. In this paper, we propose robust CEs that introduce a new viewpoint - Pareto improvement - and a method that uses multi-objective optimization to generate it. To evaluate the proposed method, we conducted experiments using both simulated and actual data. The results demonstrate that the proposed method is robust and useful. We believe that this research will contribute to a wide range of research areas, such as explainability in machine learning, decision-making, and action planning based on machine learning.

Problem

• Counterfactual Explanations
• Reliability
• Machine Learning Models

Innovation

• Counterfactual Explanations
• Pareto Improvement
• Multi-objective Optimization