Error detection rules (EDR) is a technique where rules relating to model failure can be learned (usually from the same training data as the model) and can provide a user alert as to when a model may fail. We have also shown that when these rules are learned in the multi-class case, that they can act as constraints and improve model performance (depicted below). This techniques has been applied to multiple domains.

B. Xi, K. Scaria, D. Bavikadi, P. Shakarian, Rule-Based Error Detection and Correction to Operationalize Movement Trajectory Classification. IJCAI Workshop on Spatio-Temporal Reasoning and Learning (2025).
J. S. Kricheli, K. Voa, A. Datta, S. Ozgur, P. Shakarian, Error Detection and Constraint Recovery in Hierarchical Multi-Label Classification without Prior Knowledge, 33rd ACM International Conference on Information and Knowledge Management (CIKM-24) (2024).
N. Lee, N. Ngu, H. Sahdev, P. Motaganahalli, A. Chowdhury, B. Xi, P. Shakarian, Metal Price Spike Prediction via a Neurosymbolic Ensemble Approach, Proceedings of the 2024 IEEE International Conference on Data Mining Workshops (ICDMW).