Scientific Models

I get nervous when people enthusiastically advocate models. This happens mostly, though not exclusively, in sociopolitical debates such as climate change. A scientific model is not a belief system. All those who claim that science is on their side might better consider whether they are on science’s side. Most scientific models, especially statistical ones, are ways to measure and predict the behaviour of a bounded system (a subset of nature). Such a model does not claim to be the final, fundamental, and ultimate description of nature itself. It’s important to not confuse the map with the territory, since rabbit holes rarely appear on maps.

In fact, even a major shift in the underlying paradigm may only correspond to a small change in a model’s accuracy. A very good essay on this was penned by Isaac Asimov. In short, he argued that models may be updated with more subtlety and scope, but they are not summarily discarded as simply and absolutely wrong. Long after we replaced force with space-time geometry in our gravitational models, we still design elevators with Isaac Newton looking over our shoulder. We are all very much still ‘Flat Earthers’ when we plan a walk (unless that walk will take us across a time zone boundary or we sneak a peek at our GPS location along the way).

In this view, science is models and observations only. A theory is only a specific model, not a deep insight into the true nature of reality. Nature simply is. That’s it. Our beliefs about its fundamental structure are illusions. The meaning of observations is model dependent, but the observations themselves are not. Antares is in the same position in the sky, whether it is a red supergiant or “the heart of the scorpion”.

It should be noted that not everyone agrees with this view. For example, Thomas Kuhn argued that newer models are not just ‘better approximations’ (straw man alert) and that it is a serious mistake to teach that they are. He argued that consensus of the scientific community is also a component of scientific truth. I don’t happen to share that view, but then as I consider consensus to be irrelevant in science, my opinion means naught anyway. Of course paradigm shifts and revolutions do happen, but they are artifacts of history and psychology, not objective science. It’s a fallacy to define science as being only the set of all scientists, past and present. Again, there’s that map and territory confusion.

Raw predictive power alone is not the only metric for judging a model. Probably the most universal heuristic is agreement with observation. Many a compelling model has been ruined by the facts. Mathematical beauty, symmetry, and ‘feel’ are not important. Elegance and simplicity are helpful, perhaps even necessary, but not sufficient conditions for a good model. Enthusiasm for a model is a strong cause for suspicion. Humility before nature is a much better heuristic. Darwin and Planck had to be dragged along reluctantly with their own theories. Models can incorporate ‘tricks’ that are obviously not part of reality. For example, adding an equal ‘negative mass’ allows one to easily calculate the centre of mass of a plate with a hole in it. Perhaps the best thing a model can do is to not just answer questions, but to ask even better ones. Heliocentricity and the microbe theory of disease launched tremendously fruitful new science.

Nature is a realm of computation and evolution on an unimaginable scale (quite literally). Local agency leads to emergent complexity. Mathematics is one of the tools that can allow a vastly simplified model of nature to be stored in a three-pound hominid brain. However, daisies and snails know nothing about mathematics. Anthropomorphization of nature is not science. Consensus and comfort are not science. Clinging to teddy bears is really cute in youngsters, not so much in adults. The good news is that nature is far, far, far more wonderful than anything any individual or advocacy group can come up with (including well-crafted teddy bears). You just have to let objectivity take the wheel.