Parameters, prediction, post-normal science and the precautionary principle—a roadmap for modelling for decision-making

In the wake of the ‘discovery of complexity’, dynamical simulation models have become widespread, guiding human interaction with complex systems, e.g. in ecosystem management, environmental decision-making and risk assessment. Any model establishes a reading frame for ecological phenomena or systems, determining the parameters that are assumed to be adequate for the encoding of ecological phenomena. Departing from a definition of observation as the operation of distinguishing and designating, and as the application of certain distinctions to complex phenomena, we analyze the construction of reading frames. As dynamical systems are the prevalent paradigm and reading frame for ecosystems, we describe the sequence of distinctions and selections by which scientists encode ecosystems into formal, dynamical system representations. Major shortcomings of the dynamical system paradigm are highlighted: dynamical systems are conceptually closed systems requiring a fixed set of a priori defined parameters, part of which are parameters of convenience satisfying mathematical needs and part of which are residual parameters that account for noise and system background. Ecosystems, in contrast, are conceived as conceptually open, self-modifying systems, which constantly (‘on-line’) produce novelty and new parameters and which cannot be severed from their environment. Although calibration may adapt models to data sets of the past, it does not assure predictive capacity nor validity. While models serve heuristic and theoretical functions and may outline the space of possible behaviour, they may be deficient instruments for the reduction of uncertainty as to future system behaviour. Different forms of uncertainty are at the heart of environmental decision-making, among them epistemic uncertainty, which arises when the normal, disciplinary forms of uncertainty reduction fail and which leads to debate on adequate ways of coping with uncertainty. Epistemic uncertainty in environmental issues may call for a different type of science that differs from normal, positivist science. Such post-normal science is transdisciplinary, participative and context-sensitive in that it aims at the production of knowledge for concrete, real-world problems. New forms of knowledge production such as the concept of post-normal science in conjunction with the precautionary principle challenge the established authority of science and may lead to an institutional split of science into an academic branch and a managerial, public policy branch. Correspondingly, modelling for theoretical scientific purposes and modelling for decision-making may follow separate paths. Modelling for decision-making may have to take into account requests for transparency and participation (‘deliberation frames analysis’) and the validity of model products will be judged according to their capacity of providing context-sensitive knowledge for specific decision problems.