Explainable AI models for predicting liquefaction-induced lateral spreading

Explainability tools can make AI trustworthy for critical infrastructure decisions: By using SHAP analysis to interpret a machine learning model predicting earthquake-induced lateral spreading, researchers demonstrated that AI systems can be made transparent enough for engineers to understand why the model makes specific predictions, rather than treating it as a "black box."

AI predictions aligned with established expert knowledge: The study found that the machine learning model correctly identified soil characteristics from Cone Penetration Test data as the most important factors for predicting lateral spreading, which validated that the AI system's reasoning matched what geotechnical engineers already know to be important—a key indicator that the model is reliable.

Explainability enables adoption in high-stakes applications: By revealing which factors drive predictions, explainable AI makes it easier for engineers and decision-makers to adopt machine learning for hazard assessment and infrastructure protection, where understanding the reasoning behind recommendations is essential for public safety.

Combining AI with domain expertise strengthens both: This work shows that machine learning doesn't replace engineering knowledge; instead, explainability tools allow AI and human expertise to work together, with AI capturing complex patterns while engineers verify that the model's logic makes sense.