Models and Rumours of Models

I have written a lot over time, and quite recently, about models and how we use them in wargaming. It is worth considering, I think, because wargaming, and writing wargme rules are essentially modelling activities. There is no getting away from the concepts of models at all.

Models, of course, come in various shapes and sizes. There are scale models, representational models, computational models and so on, all of which, I think, appear in most wargames. But it is also worth taking a bit of a step back and considering exactly what the models are, and what they can do.

In Chapter 3 of Ian Barbour’s Models, Myths and Paradigms (London, SCM, 1974), there is a list of four ways of understanding models in science. These are naïve realism, positivism, instrumentalism and critical realism. I will have a go at describing how these models might fit into a set of wargame rules, and which interpretations of our models we might choose.

Firstly, then, naïve realism. In this case we assume that our models give us direct access to the world. The entities we postulate in our models are those which ‘really’ exist, even if they are not directly observable. For wargaming, this means, for example, that we would have to claim that the factors we use to calculate a wargame unit’s morale were really existing factors, and that they can be summed up and reach a given outcome of ‘hold’ or ‘rout’. This is really rather, um, naïve, as the name of this view suggests. Morale calculation is a computational model, and bears little relationship to the real world.

Naïve realism corresponds to a literal view of models, that is, that the model is a replica of the world. While some of the bits of a wargame might correspond to bits of the world, for example our soldiers are scale models of real soldiers, I do not think that many wargamers would fall into the trap that our wargames are one to one readings of the real world, although some wargamers do seem to sometimes speak or write as if they are. Literalism, however, runs the risk of pushing the model too hard, of expecting, in the morale example, soldiers too be looking around and thinking ‘enemy on flank, that’s bad, taking casualties, that’s bad, rear support, that’s good. Hmm, two bads and one good, and a roll of two on the dice. That’s it, I’m running away.’ This is not a realistic model of morale in the real world.

Secondly, we have positivism. This argues that all a model is is a representation of experience, a convenient way of classifying empirical data. This claims that all we have is a means of correlating data. In wargame terms, this suggests that all we can do is give a relationship between a cause and an effect. Thus, if the effect is ‘the infantry run away’, the cause ‘the cavalry charged’ is correlated with it. Those who have good memories, or who know something about empiricism will observe the hand of David Hume sitting on this idea.

So far was wargaming goes, I think, positivism is not terribly helpful. A positivist will claim that a model has no real use, and is simply a means of abstracting from data. Thus, in this concept, we would be unable to wargame anything except historical battles. In science terms, positivism means that theories are stripped of their predictive ability. In wargame terms, then, we are limited to the strict data of history. This seems unlikely to produce a satisfactory wargame.

Thirdly, there is the idea of instrumentalism. This is similar to positivism in that the claim is made that our models are not representations of the world, are not true or false, but techniques for creating inferences. The argument here is that the terms of a model cannot be translated into observational terms, at least directly. Again, using wargame morale as an example, we can see that this might well be the case. We can assess, as a wargamer, the morale of a unit, but that assessment is not translatable into the real world directly, as I argued above. Instrumentalism, however, allows that our concepts and models may have results that are discernible in the real world. Real army units do, from time to time, run away, and the outcome of our models can predict that, even though the computational aspect of has no real world equivalent.

A model, then, in this view, is neither true nor false, neither accurate nor not. It is simply a useful way of giving us mental devices for thinking about things.  Once we understand the data and the theory behind the model, the claim is, we can discard the model; it has outlived its usefulness. In the case of, for example, wargame morale, this seems unlikely to be the case.

Finally, we have critical realism. Here, theories are representations of the world; theories are true and useful,, but incomplete and selective. Models are abstract systems, representative of some bits of the real world, created in our imaginations for a given purpose. A model, then, is an inexact account of the world, missing out huge chunks in an effort to provide something that is intelligible and tractable. Thus, a model will need validating against real world data. This, in science, is the role of experiment, while in wargaming is the role of historical accounts of battles.

It seems to me that models in wargame ruleas are something of a mix of instrumentalist models, such as calculations of morale, and critically realist models, such as combat rules. In the first, the calculation bears no relationship to the actual real world process of how units ‘decide’ to stand or run. Thus all we can do is make the calculation and compare it with empirical outcomes.

In the second case, we can actually see the process of combat in some sort of detail, and, in a highly selective manner, model it, comparing the model and its stages with the real world and its outcomes.

But, at the end of the day, we do seem to need both concepts of models to create a wargame rule set.