Domain-informed explainable boosting machines for trustworthy lateral spread predictions

2026-03-17

AI Summary

This study addresses the issue of interpretable machine learning models, specifically Explainable Boosting Machines (EBMs), learning non-physical relationships in predicting earthquake-induced liquefaction lateral spreading, which compromises their reliability. To mitigate this, the authors propose the first integration of domain knowledge into the EBM framework by imposing physical constraints on its shape functions. This approach preserves the data-driven flexibility of EBMs while enhancing their physical consistency. Evaluated on data from the 2011 Christchurch earthquake, the constrained model demonstrates markedly improved physical plausibility in both global and local explanations, with only a marginal 4–5% reduction in predictive accuracy. The results illustrate a successful balance between high interpretability and adherence to physical principles, offering a robust pathway for trustworthy, physics-informed machine learning in geotechnical applications.

Technology Category

• Machine Learning: Transparent, Interpretable, Explainable ML
• Constraint Satisfaction and Optimization: Constraint Learning and Acquisition
• Reasoning under Uncertainty: Relational Probabilistic Models

Application Category

• User Modeling, Personalization and Recommendation: Explainable and interpretable methods for personalization
• Semantics and Knowledge: Methods, algorithms and applications for the development of semantic models, knowledge graphs and other forms of structured data models with machine-interpretable semantics
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking

Abstract

Explainable Boosting Machines (EBMs) provide transparent predictions through additive shape functions, enabling direct inspection of feature contributions. However, EBMs can learn non-physical relationships that reduce their reliability in natural hazard applications. This study presents a domain-informed framework to improve the physical consistency of EBMs for lateral spreading prediction. Our approach modifies learned shape functions based on domain knowledge. These modifications correct non-physical behavior while maintaining data-driven patterns. We apply the method to the 2011 Christchurch earthquake dataset and correct non-physical trends observed in the original EBM. The resulting model produces more physically consistent global and local explanations, with an acceptable tradeoff in accuracy (4--5\%).

Problem

• Explainable Boosting Machines
• physical consistency
• lateral spreading prediction
• domain knowledge
• natural hazards

Innovation

• Explainable Boosting Machines
• domain-informed learning
• physical consistency
• lateral spreading prediction
• interpretable machine learning