Humans in the Digital Twin Loop: An Opinion

The debate around bi-directional Digital Twins (DTs) has gained traction over the last few years. While there is broad agreement on the role of a Digital Twin in representing and monitoring physical assets, there is still disagreement on what qualifies as a “real” Digital Twin. Some scholars argue that a Digital Twin must be bidirectionalwhere information flows from the physical world to the digital model and back again – actually, this was one of the comments I recieved from the reviewer who reviewd my DT definitions paper. Others take a more relaxed view, considering monitoring and synchronization capabilities sufficient, even if the Digital Twin has no direct influence on the physical asset. Although, to be fair, many of them would probably call such a system a Digital Shadow rather than a full Digital Twin.

However, I think the discussion becomes more interesting once we introduce humans into the loop. Consider a Digital Twin that is used purely for monitoring. Under the strict definition, such a system would not qualify as a true Digital Twin because information only flows from the physical asset to the digital representation. There is no actuator, no control signal, and no automated feedback mechanism that affects the physical world. But is that really the case?

In practice, one or more persons are usually looking at the dashboard. A facility manager may adjust the HVAC settings after observing an anomaly. A maintenance engineer may schedule an inspection after receiving an alert. A city operator may redirect resources based on what they see in the Digital Twin. In all of these cases, information is translated into actions that affect the physical world. The only difference is that the actuator is not a machine; it is a human or a chain of humans.

If we accept that the purpose of bidirectionality is to allow the Digital Twin to influence the physical asset, then should it matter whether the actuator is a machine or a human? From the perspective of the asset being monitored, both mechanisms produce the same outcome: information generated in the digital world results in changes in the physical world.

In fact, humans are not entirely different from other assets in this regard. They have operating limits, performance degradation, failure modes, and varying levels of reliability. They require training, monitoring, and sometimes intervention. One could even argue that humans need predictive, and occasionally, preventive maintenance.

Does bidirectionality require automation, or does it merely require that information generated by the Digital Twin can influence the physical asset?