Reality is infinitely complex. Yet business or psychology writers (and social scientists generally) want to create models of that reality. They dream of creating something perfect.
If I’m working with them, I have to tell them that this is impossible. A good model aims to capture a big chunk of that reality, do so accurately, and be relatively simple: three desiderata, that pull model-builders in three different directions.
The tension between the three (simplicity, precision and breadth) is known as Thorngates’s Impostulate, after a Canadian social psychologist, Warren Thorngate, who first came up with the idea in the 1970s.
I envisage the Impostulate as a triangle, with one the three desiderata at each corner. The job for the model-builder is to locate their model within that space. Writing for the general public, the model has to be simple. Then you start trading off precision with breadth of application. Breadth is probably the next most attractive to the general reader. A model that is simple but also seems to apply to lots of situations sounds ideal. If you can’t make precise predictions based on it – well, so what?
But some people want accuracy. Should you then move the model towards that corner, narrowing it down, and saying that it only works in certain circumstances but in those circumstances, it works very well? Good idea, but then it will appeal to fewer people.
Thorngate’s great insight was that there is no ideal answer to debates like this. You just have to do your best.
I am currently preparing a new edition of my most successful book, The Beermat Entrepreneur. In it, Mike Southon and I created a series of models. They were all simple: we wanted models that readers would remember. We are now making some of them a little more complex, to move them further towards the accuracy corner – but I’m aware of the dangers of losing simplicity. Looking back, we made some of them too general, too: the book is probably most useful in a more restricted area, business services. But we don’t want to narrow our focus down too much, as we feel that much of the Beermat approach is still widely applicable. So the new Beermat will be a little less simple, a little less general, and (we hope) much more accurate. Will this be a better model? I hope so, but genuinely don’t know.
The philosophy of science I learnt at university said that science progressed by creating models and ‘disproving’ them. But it doesn’t. If a model doesn’t work in a certain situation, this doesn’t ‘disprove’ it, but shows that it needs to be moved around inside the triangle a bit. A long, unbroken series of refutations does spell trouble for the model, but it can even wriggle out of that. Some models may retreat into the accuracy corner, to cover only a few situations, just as a stopped clock is right twice a day. Others can go the other way, and end up as empty political slogans. But most will adjust to refutation and find a compromise, a new place within the triangle.
And no model covers everything, not even the most complex and well-tested academic one. I find that a comforting thought.