A three-Level Framework for LLM-Enhanced eXplainable AI: From technical explanations to natural language

TL;DR

The paper addresses the need for audience-centered explainability in AI, proposing a multilevel framework that layers algorithmic fidelity, human-centered interaction, and societal transparency. Large Language Models serve as mediators that translate technical explanations into accessible, contextual narratives across levels while preserving the integrity of underlying XAI methods. Through loan-approval case studies, the authors demonstrate how SHAP, interpretable models, and prototype/counterfactual explanations can be communicated via LLM-driven dialogue to enhance trust, accountability, and public understanding. The work highlights regulatory relevance (e.g., EU AI Act), privacy considerations, and the importance of iterative, user-driven refinement, while outlining future directions for metrics, deployment, and governance. Overall, the framework reframes XAI as an evolving, socially accountable process that bridges technical rigor with broad societal needs.

Abstract

The growing application of artificial intelligence in sensitive domains has intensified the demand for systems that are not only accurate but also explainable and trustworthy. Although explainable AI (XAI) methods have proliferated, many do not consider the diverse audiences that interact with AI systems: from developers and domain experts to end-users and society. This paper addresses how trust in AI is influenced by the design and delivery of explanations and proposes a multilevel framework that aligns explanations with the epistemic, contextual, and ethical expectations of different stakeholders. The framework consists of three layers: algorithmic and domain-based, human-centered, and social explainability, with Large Language Models serving as crucial mediators that transform technical outputs of AI explanations into accessible, contextual narratives across all levels. We show how LLMs enable dynamic, conversational explanations that bridge the gap between complex model behavior and human understanding, facilitating interactive dialogue and enhancing societal transparency. Through comprehensive case studies, we show how this LLM-enhanced approach achieves technical fidelity, user engagement, and societal accountability, reframing XAI as a dynamic, trust-building process that leverages natural language capabilities to democratize AI explainability.

Figures (5)

• Figure 1: Diagram showing different audience profiles (inspired by figure 4 in herrera2025reflections).
• Figure 2: Algorithmic and Domain-Based Explainability level.
• Figure 3: Human-Centered Explainability level.
• Figure 4: Understandability and dialogue level.
• Figure 5: General overview of our framework for the case studies of approving a loan.