Decision Making / 4 / 4.9 / 4.9.9

4.9.9 Tame Vs. Wicked Problems—Replicability

The same tame problem may repeat itself many times. Every wicked problem is essentially unique.

Because each wicked problem is likely to be different to anything we have met before, we have to be careful about making assumptions, applying those assumptions to the problem situation, and relying on them during strategy formulation. To be safe, we should expect every new problem we confront is wicked until we can conclude otherwise.

Meadows (1989) identifies some 17 common assumptions that are problematic in what she calls the ‘current industrial paradigm’. She says these are:

‘… partly or wholly false, that they are implicit or explicit in virtually all public discourse, that they give rise to much of the persistent counterproductive behaviour of individuals and institutions, and that the harm done by them is incalculable.’

Erroneous or ill-founded assumptions that reduce our opportunities for effectively addressing wicked problems include:

• One cause produces one effect.
• A problem does not exist nor is it serious until it can be measured.
• Relationships are linear, non-delayed, and continuous; there are no critical thresholds, feedback is accurate and timely; systems are manageable through first-order negative-feedback-loop thinking.

Forrester (1985), Lane (1993), Morecroft and Sterman (1994), Richardson and Pugh (1981), Sterman (2000), Vennix (1996) and Wolstenholme (1985; 1990; 1992) are strong advocates of modelling as an important enabler to understanding and learning. They put forward compelling arguments to support the hypothesis that a highly effective way of improving our understanding of [unique, wicked] complex, dynamic problems and to promote learning is to model them. In modelling as learning (Lane, 1993; Morecroft and Sterman, 1994), cycles of virtual world modelling and simulation are used to test assumptions and build understanding of complex, dynamic problems through double-loop learning (Argyris, 1982).

Modelling and simulation even in qualitative terms needs to be done with care and scientific rigour (Homer, 1996) for learning to occur without error. Verification and validation must be conducted, as far at it is possible. Practical difficulties of validating and evaluating models mean there is always room for improvement.

Some important information that we need will be in error, irrelevant, missing, slow to emerge, or deliberately withheld. We may need to move forward with the existing model and information and accept a lower level of confidence until such time as better information becomes available.

It is not readily ascertainable, nor is it always clear that models are valid or invalid. They may even be based on incorrect data. However, these difficulties can be overcome through the application of scientific methods, building quality ‘fit for purpose’ models and continual iterative cycles of validation.