Modelling Research

Models are developed and used to help us (scientists, engineers, decision makers) to understand and communicate about a system of interest with the ultimate aim of bringing a positive change to how a system is built and/or managed. By definition, models are inherently wrong.

First, they are and must be a simplified representation of the modelled system. To be useful, models need to provide a cognitively-mediated environment to explain the systemic behaviour. If a model keeps growing in complexity, it will be difficult to understand and use. Yet, in many applications, large models are inevitable to support system understanding, and decision making at the appropriate and acceptable level of detail This has been long recognized as the 'modelling paradox' (Bonini, 1963). Second, models are limited to the collective cognitive complexity of the "mental models" of those involved in model development. Our mental models (another type of models) are a flawed, incomplete, and sometimes inconsistent, representation especially of dynamic and complex systems (Rouse and Morris, 1986). These challenges are escalated on using models in deep uncertainty situations where stakeholders do not know or cannot agree on the system structure, future scenarios, and model parameters (Kwakkel and Pruyt, 2013).

In an increasingly dynamic and complex world, we therefore need to develop and use models in novel and cost-effective ways that improve our decision-making capacity. This raises the following questions:

  • How can we develop modelling frameworks and techniques that allow for systematically and automatically generating large ensembles of modelling scenarios, including combinations of various modelling hypothesis, structure, and parameters?
  • How can we make use of analytical, simulation, and hybrid techniques to examine the robustness of decision outcomes under different scenarios and objectives?
  • How can we develop hierarchical models that can be expanded and collapsed to communicate about the model structure at different abstraction levels for different audience and communication needs?
  • How can we make use of big data sets and data analysis techniques to help infer model structure?
  • How can we implement and use (all of the above) to improve decision making in problem cases characterized with complexity and deep uncertainty? What are the opportunities cross-application learnings to improve the know-how, transferability, and reduce the development costs?

Charles P. Bonini (1963) Simulation of information and decision systems in the firm, Englewood Cliffs, N. J.: Prentice-Hall

Kwakkel, J. H., & Pruyt, E. (2013). Exploratory Modeling and Analysis, an approach for model-based foresight under deep uncertainty.Technological Forecasting and Social Change, 80(3), 419-431. Rouse, W. B., & Morris, N. M. (1986). On looking into the black box: Prospects and limits in the search for mental models. Psychological bulletin, 100(3), 349.