Life lessons from and for distributed MPC – Part 1: Dynamics of cooperation

This paper and a second accompanying paper (Olaru et al., 2018) explore the potential of Distributed Predictive Control (DMPC) literature to provide valuable insights into social behaviour. In particular this first paper focuses on the mechanisms of group regulation in social systems. It will be noted that there are major differences between the way in which DMPC algorithms and Social Human Participants (SHPs) form decisions. DMPC can make optimal decisions but these are only optimal with respect to a given objective and model, both of which must be explicit. SHPs operate, by and large, with only vague, implicit objectives and models – which can be surprisingly accurate – but often make sub-optimal decisions both individually (because of irrationality or poor anticipation and due to a short horizon, bad model or misjudgement of objectives) and in a group sense (for the previous reasons plus selfishness). Thus while SHPs’ decisions would typically be suboptimal, with respect to their desired goals, for the aforementioned reasons, it can be expected that SHPs’ decision making would evolve towards an optimal solution as groups of SHPs develop more experience within the system they’re operating in.