Hybrid Human-Algorithm Committees In Production Systems: Conceptual Framework and Computational Application To Predictive Maintenance

Contemporary production systems require reliable decisions under uncertainty in tasks such as predictive maintenance and operational monitoring. This paper proposes modeling such decisions as the outcome of hybrid human–algorithm committees that combine machine learning models, human judgment and governance-related signals and coordination mechanisms within a unified deliberative framework. A computational case study based on the NASA C-MAPSS turbofan degradation benchmark operationalizes this idea by integrating multiple classifiers, a synthetic human policy agent and committee aggregation mechanisms using ensemble optimization, stacking and selective delegation. The experiment compares isolated models, an optimized algorithmic committee and the proposed hybrid committee. Results on the official test set show that the hybrid committee achieves the best balanced predictive performance among the evaluated approaches under F1-oriented calibration, reaching an F1-score of 0.7319 while maintaining high accuracy and strong discrimination capability.