Smart Manufacturing and the Operator’s Digital Double: Modeling Cognitive Load Through a Psychosocial Digital Twin

Abstract

Digital twin technology, a virtual model that replicates real-world machines, has become a key component of modern manufacturing, enabling companies to predict problems before they occur and enhance operational efficiency. Yet, most of these systems are built around equipment, overlooking the human operators who play a crucial role in the production process. To address this gap, we propose the Psychosocial Digital Twin (PDT), a framework designed to create a real-time virtual model of a worker’s cognitive state. Unlike traditional monitoring tools, the PDT combines multiple data sources to track and predict stress and workload as they unfold.

To test this idea, we created a virtual factory environment using VR and conducted an experiment with 70 experienced factory workers. Participants were split into two groups: one used the new PDT system, while the other relied on conventional monitoring methods. The PDT combined information from several streams, including machine performance data (such as speed and error rates), environmental conditions (like noise and lighting), and non-invasive physiological measures (such as heart rate variability, electrodermal activity, and eyetracking). All of this was processed by an AI model that produced a Cognitive Load Index (CLI), a score showing the worker’s real-time mental stress levels. Supervisors in the PDT group could then run “what-if” simulations to test how proposed changes might affect workers before applying them on the floor.

Results demonstrate that the PDT enhanced both worker experience and operational stability. The system predicted stress events with 87.4% accuracy, reduced reported stressful episodes by 42%, and cut task-related errors by 28% compared with the control group. Supervisors also proactively altered or canceled 65% of stress-intensive tasks based on simulations.

Overall, the PDT represents a shift from reactive human factors analysis toward proactive, simulation-driven design. This study contributes to understanding human behavior in cyber-physical environments by modeling how cognitive load dynamically influences performance and decision-making in AI-augmented workplaces. By making workers’ wellbeing visible, measurable, and optimizable, this framework provides a scalable method for balancing productivity and safety, thereby enhancing performance in Industry 4.0 environments.