Mervyn King is under fire for (amongst many sins, real and perceived) for being misled by ‘faulty’ models of our economic system. Apparently a number of former members of the MPC have agreed to join an ‘extraordinary’ experiment with a consultancy ‘Fathom’ (could be a lot of them so take this link with a pinch of salt) to develop an alternative forecasting model. Of course they expect this will be better.....

Let’s face it, this is rubbish.

How can a single ‘model’ (it’s not reality folks, just a map) be understood? With an enormous number of parameters, all of which can be ‘tweaked’, all of which influence other parts of the system (can anyone spell non-linear?) is it possible for anyone to really understand the ‘model’ and if anyone really does (that one very lucky individual) should we leave ourselves vulnerable to any single interpretation?

Fundamentally there are two extremes in model creation - the structural and the phenomenological. At one end, the model makers are interested in exploring the structure of the systems they wish to understand, comparing and contrasting observations with the predictions that their models make. At the other, observations drive the models and ‘curve fitting’ - models that just try to imitate observations without necessarily taking any account of the ‘facts on the ground’.

Now these are ridiculous extremes - most models live somewhere in between, a combination of structural understanding (or assumptions) and ‘where we don’t know - fit whatever can be measured’. Fair enough, you do what you can.

What many users of models (and their creators) don’t appear to understand (at a visceral level) is that their models are not reality. What they are (or should be interpreted as) is a set of windows into reality, each one of which allows the viewer to see and understand a very particular part of the landscape.

Terry Pratchett, in a very amusing book Making Money, plays games with models and their relationships with reality. His ‘Glooper’ (based we assume on the Phillips Hydraulic Computer) gets a little entangled with reality. The Glooper, intended to model the economy of Ankh-Morpork begins to control it. Change the model, and reality shifts.

This is silly. But however silly it might seem, too many organisations, including the MPC seem to get the two - model and reality a little confused. It takes a novelist to say something sensible. Models need context (their ‘stakeholders’) and anything that can not be understood should not be used. In fact organisations that depend on any prediction that cannot be discussed, understood, criticised and refined are putting themselves at unnecessary risk.

There is an alternative approach to the ‘grand unifying theory’ (which brought down Einstein in his later years) and that is the KISS principle. Keep the models simple, keep the models accessible, keep your stakeholders connected and keep the conversations going. Don’t look to a single model to explain your system, rather craft smaller, directed models that are intended to address specific questions. This means that you can carry on your conversations with the many people who can not only help with the structure and the parameters, but you can also make use of different modelling technologies that are most appropriate to the questions being posed.

Organisations do seem to get confused about the difference between the map and the territory. If we don’t want to be ‘gloopered’ again, it might be worth while remembering that.

Caroline Bosworth {caroline.bosworth@concinnitas.co.uk}