Human fatigue and recovery are fundamental yet poorly quantified processes governing performance and safety in labor-intensive and hybrid human-machine systems. Despite decades of ergonomics research, current models cannot predict fatigue evolution or recovery across varying work conditions, particularly when human performance degradation and restoration occur simultaneously. This study defines the analytical and methodological challenges of modeling human fatigue and recovery and demonstrates, using evidence from a controlled order-picking experiment, the need for continuous, data-driven representations of competing fatigue and recovery dynamics under uncertainty. The analysis highlights key barriers including nonlinear dynamics, multimodal and asynchronous data, uncertainty quantification, interpretability, fairness, and privacy that constrain existing reliability and machine learning frameworks. Addressing these barriers opens new research opportunities in Industrial and Systems Engineering (ISE), including reliability theory, hierarchical and causal learning, uncertainty-aware decision analytics, and ethical data infrastructures for human digital twins. Collectively, these directions articulate a convergent ISE vision with human-centric analysis and decision-making systems in which humans, AI, and machines co-evolve, adapt, and respect each other’s limits to achieve resilient, safe, and ethically grounded intelligent operations.