Mind the XAI Gap: A Human-Centered LLM Framework for Democratizing Explainable AI

The "XAI Gap" is a real problem: Current explainable AI tools are designed primarily for experts, leaving non-experts unable to understand how AI systems make decisions that affect them—a critical issue since AI is increasingly used in important decision-making systems.

One framework can serve multiple audiences: The paper demonstrates that a single explanation system can simultaneously provide accessible explanations for non-experts and technical details for specialists, rather than requiring separate tools for each group.

LLMs can bridge the explanation gap: By using Large Language Models with structured prompts, the framework translates complex AI explanations into human-friendly language while maintaining technical accuracy—achieving strong correlation (0.92) with expert-validated explanations.

Human-centered design matters for AI transparency: The framework prioritizes what non-experts actually need to understand AI decisions, not just what experts think should be explained, addressing the "unprecedented human values risks" that come with AI in critical systems.

The solution is generalizable across different AI systems: Rather than building custom explanations for each AI model and domain, this framework works across different types of data, models, and explanation methods, making responsible AI more practical to implement broadly.