Managing Epistemic Uncertainty for Multimodels of Sociotechnical Systems for Decision Support

Multimodeling has become an increasingly popular way to analyze sociotechnical systems for decision support. Just as with physics-based multimodels, one would like to be able to quantify and manage uncertainty associated with the model to support decision making. However, when we consider multimodels of sociotechnical systems, epistemic uncertainty with regard to the structure of the model itself erodes their effectiveness. In this paper, we characterize these uncertainties as consequences of bifurcations that result from the complexity of sociotechnical systems. One could try to compensate by increasing the fidelity of the model, but this may come at the cost of actually increasing both the epistemic uncertainty and the computational burden. In short, building an extremely detailed model of a complex system may not be worth the effort. We conclude that in many cases, it may be better to build a low-fidelity multimodel of a sociotechnical system and use it to identify the existence of bifurcation points that can be managed through adaptive and hedging strategies. This is in contrast to building a very detailed model that endogenizes the epistemic uncertainties to support optimization. The latter approach is likely to be effective for physics-based systems but not for sociotechnical systems.