The evolution of Digital Twins is towards the acquisition of an autonomous role, i.e. a digital twin may start to act as an autonomous entity performing task on behalf of its physical twin. Although this is a natural evolution as digital twins shift from a static copy to a dynamic aggregation of services customised for its physical twin, there is also no doubt that this evolution changes the nature of the idea of a digital twin. New issues are coming up, like:
- the divergence of the digital twin from its physical twin. By taking up activities on behalf of its twin it accrues experiences that are unique to the digital twin and most likely unseen by its physical twin;
- by taking autonomous actions, on behalf of its physical twin, raises the issue of accountability and responsibility;
- acting of behalf of the physical twin can play at several levels of responsibility. A digital twin detecting a potential issue in its physical turbine could autonomously decide that a replacement of lubricant is needed and can search the enterprise resources to locate the lubricant and the maintenance crew to execute the replacement. Can also interact with digital twins of other turbine to “understand” the impact of the lubricant used by different turbine versus the actual work and stress situation to take a decision on what could be the best choice for replacing. This choice may not be available within the enterprise and the digital twin may “order” the appropriate lubricant from a seller…. It is easy to complicate at will the implication of operation in an autonomous space.
Now, all of this gets even more awkward when we are dealing with a Personal Digital Twin. The example of the creation of a personal digital twin by UBS to clone their Chief Economist is a point in case. The real Chief Economist, Daniel Kalt, can only talk to a very limited number of clients every day, by creating its personal digital twin/avatar, UBS –and IBM Watson- is able to share the knowledge of Daniel to a multitude of clients and in doing so, through clients interactions the PDT acquires information that is not available to Daniel. It can also check, analyse that information with thousands of data available on the Internet, learn from correlation and quickly get “better” in some sense than Daniel. Now, this is a crucial point: what does it mean to get better than Daniel? In general it would mean for the client to receive more actionable and accurate advices resulting in a gain (or limiting losses) and for the bank to make more money from consultancy services (the two may not necessarily go hand in hand). Would the real Daniel, if he was aware of all the info available to his PDT react in the same way? If not, how can we say that Daniel is accountable for his PDT? And if he is not accountable, who is? IBM Watson, UBS…? (Don’t worry on the specifics, UBS and IBM have plenty of lawyers that crafted documents for limiting their liability…).
Getting back to the use of PDTs in the context of epidemics, it is natural that having an autonomous PDT could in principle increase its effectiveness. The PDT could interact with other PDTs creating a local awareness of what is going on, without having to rely on information coming from “above”. This might also be a way to preserve higher level of privacy, since data would only be shared among a limited number of PDTs and that information will fade away as it will no longer be needed. At the same time, security issues, that are however present in all aspects of data use and processing in the frameworks previously discussed, become crucial. How can a PDT trust another PDT and to what extent can data be shared with confidence? Blockchain may be of help, as well as reputation mechanisms.
Once a person has overcome Covid-19, infected and recovered, she is no longer at risk of re-infection (at least this seems the general scientific agreement) nor can she be a source of contagion for others. Hence her PDT can advise other PDTs that may get, or plan to get in her vicinity that there is no risk. Conversely, a person that has been in the vicinity of another person that turned out positive to Covid-19 can have his PDT signalling the potential risk to others. Again, this approach can better preserve privacy than sharing this information at global level.