Explaining Complex ML Models to Domain Experts Using LLM & Visualization: An Exploration in the French Breadmaking Industry
Résumé
Modeling a complex system from data can aid understanding and decision-making. Bayesian networks are one such method that, when accurately constructed, can support inference and help understand the underlying system that generated the data. However, the outputs of these models are not always intuitive, especially for users that lack a statistical background. In this work, we examine how the recent advancements in modern Large Language Models (LLMs) may be applied to help explain machine learning (ML) models. Following a user-centered design methodology, we collaborated with a team of ML modelers and a domain expert in the French breadmaking industry to develop a causal inference application with an integrated chat assistant. From qualitative feedback sessions with modelers and the domain expert, we note some unique advantages but also a host of challenges in using current LLMs for model explainability.
Domaines
| Origine | Fichiers produits par l'(les) auteur(s) |
|---|---|
| Licence |
