12,445 papers · continuously updated · last export: 10 Aug 2026livingmeta.ai
← Browse all papers
AI evidence extraction

LLMs for Explainable Artificial Intelligence: Automating Natural Language Explanations of Predictive Analytics Models

Sonsoles López‐Pernas, Yige Song, Eduardo Oliveira, Mohammed Saqr · 2025

AI-generated evidence extraction, verified across multiple analytical personas. Not a substitute for the peer-reviewed original.

5/10
Relevance
0/4
Quality (LMQS)
D
Evidence
2
Citations
6.65
FWCI
Top 10%
Impact

This is an AI analysis. Read the peer-reviewed original at the publisher: https://doi.org/10.1007/978-3-031-95365-1_11

Methodology & findings

Study design

Conceptual demonstration using open source language model through LM studio software; illustrative case study approach showing integration of LLMs with XAI outputs

Main result

The paper demonstrates that "LLMs can contextualize feature importance, partial dependence profiles, and local explanations, making predictive model outputs more interpretable and actionable for non-technical stakeholders" through integration into the XAI pipeline within learning analytics.

Reports effect sizes.

Research paradigm

interpretivist/constructivist

Author conclusions

The authors conclude that "this work contributes to advancing the intersection of generative AI, XAI, and learning analytics to promote transparency, inclusivity, and fairness." The integration demonstrates the potential of LLMs to make predictive model outputs more interpretable for non-technical stakeholders through natural language generation.

Risk of bias

No empirical validation or evaluation metrics reported; Single example implementation without comparative analysis; No user studies or stakeholder feedback; No empirical validation or user studies conducted; Single model implementation demonstrated without comparative analysis; No assessment of explanation quality or user comprehension; Limited scope - conceptual paper without experimental validation

Data: not_statedCode: not_statedExtracted from: pdfAgreement 76%

Explore related topics

Related papers