LLMs for Explainable AI: A Comprehensive Survey

LLMs can translate "black box" AI into human language: The paper identifies a key problem—sophisticated AI models like neural networks are difficult for users to understand—and proposes that Large Language Models can help by converting complex model outputs into clear, narrative explanations that non-experts can comprehend.

Explainability is essential for trust and accountability: According to the survey, when users cannot understand how AI models reach their conclusions, it undermines trust in AI systems, reduces accountability, and makes it harder to identify potential biases—highlighting why explainable AI matters for responsible deployment.

There are significant challenges in evaluating LLM-generated explanations: The paper emphasizes that while LLMs show promise for creating explanations, there are substantial challenges and limitations in assessing whether these explanations are actually accurate, helpful, and trustworthy—suggesting this remains an unsolved problem.

A multidisciplinary approach is needed going forward: The survey calls for future XAI work via LLMs to be more "interpretable, automated, user-centric, and multidisciplinary," indicating that responsible AI explainability requires collaboration across different fields and perspectives, not just technical solutions.

Real-world applications exist but need careful implementation: The paper examines actual uses of LLMs for explainability, suggesting the technology is moving beyond theory, though the emphasis on challenges implies practitioners should proceed thoughtfully when deploying these systems.