Many societal, environmental and technological challenges can be characterized as wicked problems by virtue of being difficult to understand, define and solve. examples include sustainable management and consumption of resources, resilient technical infrastructure or curbing plastic pollution of the oceans. One method of tackling such wicked problems is the use of computer-aided modelling and simulation. Model-based decision support is a growing discipline involving the use of computer models of complex systems to explore, understand and manage them. A core concept in model-based decision support is scenario discovery. In scenario discovery, a model’s inputs and outputs are related to understand under which conditions policy-relevant outputs may occur. In a first step, a diverse set of inputs is used to generate a variety of outputs. In a second step, the subset of decision-relevant outputs is identified among the outputs
through some external criterion, such as a threshold value. Finally, the inputs which generated those outputs of interest are identified, and a generative rule set is induced which usefully predicts under which conditions an input will generate an output meeting the external criterion. This rule set bounds an input subspace of interest, from which (most of) the outputs of interest originate. While scenario discovery performs adequately for quasi-linear and simple models, it is not well suited to behaviorally complex, nonlinear models. This is both because external criteria are hard to define for complex model behaviors, and also because there are often significant interactions and dependencies between model inputs, which current rule induction algorithms have trouble identifying.......