All models are wrong, some models are useful
While the statistician George Box is generally attributed with the statement, “All models are wrong, but some are useful”, my first encounter with it was when I was starting in a coaching role at Driven Alliance and was in a informal training session with Danie Roux & Kevin Tretheway.
At the time we were looking at various models for understanding and predicting team behavior. Danie & Kevin were trying to get across as strongly as they could that while the model we were looking at was useful (if I remember right it was Tuckmans model), it was fundamentally an over-simplification and thus wrong.
As they explained, all models are simplifications - that’s why they are models. The model makes it easier for a person to grasp certain concepts about the thing being modelled without getting overwhelmed by the complexity of the actual thing HOWEVER with the model being a simplification, it is also fundamentally wrong (there will always be instances where the model doesn’t explain or predict the behavior correctly). Make sense?
Models are useful when they are kinda right
When is a model useful? Originally I thought models were useful when they were correctly explaining or predicting behavior.
For instance, with Tuckmans model, I find it useful when I can look at a newly formed team and see that they are still “Norming”. Being able to predict that the team will transition from a norming to a storming phase is also extremely useful. While I still think models are useful when they are explaining or predicting things correctly, I’ve got a new angle on when a model is useful.
Some models are useful because they are clearly wrong
Some models are useful because they are clearly wrong. Yes, I know this may sound counter intuitive - let me explain why…
I really battle to come up with new ideas. Let’s say I’m working on some marketing material - for instance, a brochure for a product. I can sit for hours and not come up with anything new. My brain just seems to battle with this process.
Show me a ‘proposed’ brochure and within seconds I can tell you what I would change on it. Having the ‘wrong’ thing becomes useful to getting to the ‘right’ thing. I think the same applies with models.
One of my favorite talks at software conferences this year was a talk done by Sarah Mei. I watched it at Ruby Fuza in Cape Town, and then a revised edition at CodeMania in Auckland. Both times I really enjoyed it.
Sarah’s talk was about how we make software and the models we use to explain team software development. She proposed a new model called the stage model. In getting to her stage model, she first went through two widely accepted models, the Factory Model and the Craftsmanship Model of Software Development.
While exploring these two models, she pointed out why each model was wrong and why it did not answer the questions she had about team based software development. Had she merely jumped to her stage model without exploring the flaws in the other models, I don’t think I would have appreciated why the stage model was an improvement. Being able to understand what was wrong about the models gave me a deeper understanding of the thing being modeled.
Next time you come across a model that appears to be wrong. Appreciate it. It may be enabling you to get to a better model that gives you a deeper understanding.
All models are wrong. Some models are useful because they are wrong!