Human Digital Twins: Two-Layer Machine Learning Architecture for Intelligent Human-Machine Collaboration

Systems around us, either commercial, industrial, or social, are rapidly becoming more complex, more digital, and smarter. There is also an increasing conviction that the effective management and control of various scenarios in such complex systems can only be achieved by enabling an intelligent collaboration between the involved humans and machines. A major question in this area is how to provide machines with access to human behavior to enable the desired intelligent and adaptive collaboration between them. Based on the industrial concept of digital twins, this study develops a new approach for representing humans in complex digital environments, namely, a human digital twin (HDT). The HDT is a smart machine that learns the behavior of a human in terms of his/her communication patterns with the smart machines he/she interacts with in a specific scenario. The learned patterns can be used by the HDT and other machines supporting a human to predict human-machine (H-M) interactions outcomes and deviations. Unlike current approaches, the HDT does not need to rely on the content of the H-M interactions to learn patterns and infer deviations, it just needs to register the statistical characteristics of the exchanged messages. Using HDT would enable an adaptive H-M collaboration with minimum interruptions and would provide insights into the dynamics and dependencies involved in H-M collaborations which can be used to increase the efficiency of such collaboration.