Abstract
Cognitive Digital Twins (CDTs) of society is one way of how AI, simulation and human behavioural modelling will change, with high fidelity virtual models of people and populations. These cognitive, emotional and social smart digital analogs are not only a simple extension of digital twins, they have cognitive, emotional and social capabilities that enable them to dynamically simulate emerging human decisions, interactions and patterns in the society. By combining the results of the multimodal approach, the enormous power of huge language models, the ability to simulate the behavior by means of agents, and real-time sensing technologies, the development of adaptive Virtual Societies becomes possible, which can predict the behavioural trends, the effects of policies, as well as the systemic risks in society. The paper elaborates on the conceptual frameworks, enabling technologies and architectural structures of cognitive digital twins that are part of society simulation.