Explainable Artificial Intelligent (XAI) for Predicting Asphalt Concrete Stiffness and Rutting Resistance: Integrating Bailey's Aggregate Gradation Method

XAI makes complex material science predictions transparent: The study used SHAP (SHapley Additive exPlanations) values to explain how a deep learning model predicts asphalt concrete performance, demonstrating that explainability techniques can help engineers and stakeholders understand why an AI system makes specific predictions rather than treating it as a "black box."

Aggregate size at 0.6 mm emerged as a critical decision point: The research identified that the 0.6 mm sieve size significantly influences both stiffness and rutting resistance in asphalt concrete, showing how XAI can surface actionable insights about which specific material parameters matter most for real-world performance.

Different aggregate sizes have specialized roles: The findings revealed that coarse aggregates primarily affect rutting resistance while medium-fine aggregates influence stiffness, illustrating how explainable AI can help practitioners understand the nuanced, size-dependent contributions of different components rather than treating aggregate gradation as a single undifferentiated factor.

Practical accessibility was prioritized alongside accuracy: The researchers developed web-based interfaces with explainable features built in, suggesting that responsible AI implementation includes making both predictions and their explanations accessible to non-technical users who need to apply the results.

Model validation strengthened trustworthiness: The study used k-fold cross-validation and compared performance against alternative machine learning approaches, demonstrating that rigorous validation practices are essential for building confidence in AI systems used for infrastructure and safety-critical applications.