2.3.3 Art And Science Of Model Building
The science of model building lies in activities such as formulating and testing dynamic hypotheses about the drivers of real-world behaviour, and comparing model outputs with the real world. Models must always be simplifications of reality. So, the art lies in choosing how to construct the model to be both a necessary and sufficient representation to be useful in informing real decision making.
As far as soft variables and intangibles are concerned, human thinking is based on fuzzy logic. We continually make adjustments for those soft variables or intangibles by estimating the effect each may have individually or in combination. Models we build can never be adapted as rapidly as we might change our minds when presented with new information. Even if we could change them rapidly, they would always be redundant and out-of-date artefacts of our thinking. They will always be high-utility devices for communicating ideas and making public what we are thinking. To that end, even models which might incorporate soft variables or intangibles must be formulated as single entities with hard formulations of variables and structural representations of interrelationships between variables. So each model, whether or not it accommodates the influences of soft variables or intangibles, must become a hard representation. Such models must present only one perspective of the problem being analysed. Whilst there may be many perspectives, and we might look at the problem from each of those many perspectives in turn, we can only look at one at a time. The current model will, therefore, always be a hard model of a single (current) perspective.
We need to be careful not to hold on dearly to a single perspective regardless of how much effort was invested in the development of the model of it. Nor should we stop modelling when we have build only one model of one perspective which we believe to be sufficiently complete. The art of modelling is closely linked to having the skill to decide which perspectives to model and how many perspectives (or views) are needed to produce a sufficient understanding of the problem under study.