This little post presents in detail my favorite quote :
"All models are wrong, some are useful."
George Box
Very often, when I present scientific work to someone, that person's first reflex will be to look for a counterexample. This is a very common reflex, one that can be counterproductive.
Modeling the world exactly is impossible. It would be great: with an exact replica of the universe that we would play in a fast motion, we could make perfect predictions. But that's impossible.
In science, researchers create many models of the world, which are approximations (except for those). Since we can't model the world exactly, these models are simplifications. And because they are simplifications, they cannot be perfect. A few examples among thousands :
- The notion of social class (in economics and sociology) is a model. For example, the poverty level is often set at 60% or 40% of a country's median income. This does not mean that an individual who earns 41% of the median income is not poor. It's an arbitrarily set threshold - a model - but it shows, for example, that there are more poor people in Germany than in France.
- Classical mechanics is a model that can accurately predict the trajectory of most bodies in our solar system (except Mercury). Even if it is imperfect, this model is very useful.
- Moore's Law in computer science predicted that the number of transistors on a processor would double every two years. Today it is no longer true, but this model has been useful for decades.
We can find like this a multitude of models that are approximately true. But if one is aware of their limitations, then one also knows when they are useful. There is no point in systematically dismantling them.