Leveraging Complementary AI Explanations to Mitigate Misunderstanding in XAI

2025-03-01

AI Summary

This work addresses semantic misinterpretation of model explanations by users in eXplainable Artificial Intelligence (XAI). To mitigate ambiguity, we propose a complementary explanation paradigm: generating explanation pairs—comprising a primary explanation and a semantically aligned auxiliary explanation—that proactively resolve potential ambiguities while ensuring semantic coherence and eliminating redundancy. Methodologically, we integrate cognitive modeling with explanation consistency constraints, jointly optimizing for user misconception propensity, semantic coverage quantification, and redundancy detection to enable coordinated generation of explanation pairs. We further introduce the first evaluation framework for complementary explanations that jointly incorporates qualitative principles and quantitative metrics. Experiments across multiple XAI benchmark tasks demonstrate that our approach reduces user misinterpretation rates by 42% on average, significantly enhancing explanation credibility and decision consistency.

Technology Category

• Humans and AI: Explainable AI (XAI) for Human Understanding
• Computer Vision: Interpretability, Explainability, and Transparency
• Philosophy and Ethics of AI: Accountability, Interpretability & Explainability

Application Category

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

Abstract

Artificial intelligence explanations make complex predictive models more comprehensible. Effective explanations, however, should also anticipate and mitigate possible misinterpretations, e.g., arising when users infer incorrect information that is not explicitly conveyed. To this end, we propose complementary explanations -- a novel method that pairs explanations to compensate for their respective limitations. A complementary explanation adds insights that clarify potential misconceptions stemming from the primary explanation while ensuring their coherence and avoiding redundancy. We also introduce a framework for designing and evaluating complementary explanation pairs based on pertinent qualitative properties and quantitative metrics. Applying our approach allows to construct complementary explanations that minimise the chance of their misinterpretation.

Problem

• Mitigate misinterpretations in AI explanations
• Propose complementary explanations to clarify misconceptions
• Introduce framework for designing and evaluating explanation pairs

Innovation

• Complementary explanations mitigate AI misunderstandings
• Framework designs and evaluates explanation pairs
• Ensures coherence and minimizes misinterpretation risks