PONTE: Personalized Orchestration for Natural Language Trustworthy Explanations

TL;DR

PONTE is a human-in-the-loop framework that personalizes natural language explanations in AI, improving trustworthiness and alignment with user preferences through iterative validation and adaptation.

Contribution

It introduces a novel personalized, verification-driven approach for generating trustworthy explanations tailored to individual user needs in XAI.

Findings

Significantly improves explanation completeness and style alignment.

Achieves high agreement between user preferences and generated explanations.

Demonstrates robustness and positive quality in healthcare and finance domains.

Abstract

Explainable Artificial Intelligence (XAI) seeks to enhance the transparency and accountability of machine learning systems, yet most methods follow a one-size-fits-all paradigm that neglects user differences in expertise, goals, and cognitive needs. Although Large Language Models can translate technical explanations into natural language, they introduce challenges related to faithfulness and hallucinations. To address these challenges, we present PONTE (Personalized Orchestration for Natural language Trustworthy Explanations), a human-in-the-loop framework for adaptive and reliable XAI narratives. PONTE models personalization as a closed-loop validation and adaptation process rather than prompt engineering. It combines: (i) a low-dimensional preference model capturing stylistic requirements; (ii) a preference-conditioned generator grounded in structured XAI artifacts; and (iii) verification modules enforcing numerical faithfulness, informational completeness, and stylistic alignment, optionally supported by retrieval-grounded argumentation. User feedback iteratively updates the preference state, enabling quick personalization. Automatic and human evaluations across healthcare and finance domains show that the verification-refinement loop substantially improves completeness and stylistic alignment over validation-free generation. Human studies further confirm strong agreement between intended preference vectors and perceived style, robustness to generation stochasticity, and consistently positive quality assessments.