<p>This research presents an interpretable deep learning system that uses a Gated Recurrent Unit (GRU) with a Feedforward Neural Network (FNN) augmented with SHAP (SHapley Additive exPlanations) to increase transparency in AI-supported education extension systems. The GLU–FNN hybrid model predicts student adaptability using the OULAD dataset, evaluating both sequential learning behaviors and static demographic information. The GLU extracts sequential features that are concatenated with static representations generated by the FNN prior to classification, allowing the model to jointly learn dynamic and contextual patterns of learning. The data was split into an 80% training and a 20% testing split while fivefold cross-validation was used during model implementation in order to emphasize generalization. To address class imbalance, stratified sampling was used as the model was evaluated. The model performed strongly, achieving 0.9667 accuracy, 0.9683 precision, 0.9667 recall, and 0.9662 F1-score. The SHAP analysis was used to look into both the temporal (behavioral) and static (demographic) features, demonstrating how sequential learning trends affected adaptability predictions. While these results would suggest that explainable models may enhance educators' understanding and trust in the AI-supported learning analytics process, the claims of teacher trust as a signal of impact were made in that they are potential impacts which require empirical research and further examination.</p>