Simulating Human Behaviour: the Invisible Choreography of Self–Referential Systems

Challenges arise when it comes to capturing complex patterns of human behaviour in agent-based simulations. For example, human beings are not limited to one identity, to local levels of awareness or to acting on predetermined rules. Seemingly immune from these difficulties are some self-referential problems – situations where agents’ forecasts act to create the world they are trying to forecast. The emergent complexity in these systems results more from ways in which agents interact and react rather than from their individual idiosyncrasies. A well-known example is the Bar problem, whose collective regularities are relatively insensitive to the vagaries of individuals. Some other sociotechnical, socio-economic and socio-ecological systems of a self-referential nature are discussed. Many self-referential systems are intriguing because there is an air of inevitability about them. They seem to co-evolve in prearranged ways, as if under the spell of an invisible choreographer.