All Models Are Wrong

“Objective truths established by repeated experiments that give consistent results are not later found to be false.”

— Neil deGrasse Tyson, Starry Messenger

I reread the sentence a few times. And then set the book down on my uncle’s coffee table to think it over. I’ve always loved science and the way it allows you to organize and test ideas about the world. But I badly wanted to find a pencil to write ^often or ^usually after “not” in Tyson’s statement in my uncle’s book. Tyson lists some facts that he considers objectively true. I would be dumbfounded to learn anything in that list was scientifically disproven in my lifetime. However, there are clear historical examples of experimentally grounded scientific consensus being wrong. Ether and Caloric don’t exist despite holding consensus positions for a very long time. Tyson might say these ideas weren’t wrong. They were just incomplete, as he argues with Newton’s ideas. But Ether and Caloric don’t exist. Caloric theory wasn’t incomplete; it was false.

If the purpose of science is to generate objective, irrefutable truth, it has a mixed track record at best. However, science is fantastic at generating models with immense explanatory power. The atom model is my favorite example. Is the Bohr model of the atom “true” or “false”? I’m aware that it’s been superseded by some Schrödinger quantum probability field model, but when I think of an atom, I picture the Bohr model. Is it inaccurate in some ways? Yes. But it’s still useful. In some cases, the Lewis Dot Structure is an even better model because it’s more useful! If you want some idea of how electronics work, the Bohr model gives you little electrically charged spheres that can easily separate from the rest of the atom and go flying off to have some fun. It even gives you a nucleus that can split apart and generate incredible amounts of energy. Probability field clouds are overly complex for most purposes. It’s much harder to imagine a probability field running through copper cables. No one ever thought that a Lewis Structure was the most complete model of an atom. It’s clearly abstracting away a ton of accurate information about the atom. You might even say it’s false or wrong. It shows sodium with one electron when it actually has eleven! However, it shows exactly the information that is most helpful for combining atoms into molecules. If you want an intuitive understanding of why one oxygen atom pairs so nicely with two hydrogen atoms, it just can’t be beat.

Everyone in the world is engaged in the pursuit of truth. We engage in self-deception at times, but our brains are inherently truth-seeking machines. They want to know the truth about the temperature in the room, the weather outside, and the meaning of life. But there’s a difference between truth-seeking and objective-fact-seeking. If my thermostat tells me my house is 70 degrees, but it’s actually 71, it may have given me an objectively incorrect fact. Does it bother me? Not in the slightest, because it gave me a useful number that I can act on. I don’t need to turn on the heat. That’s what I needed to know. My brain is searching for models that describe reality well in the ways which are actionable. I’m not claiming this is the best definition of “truth” but I do think it’s the thing we actually want most of the time. We want something we can understand that will allow us to make informed choices that are aligned with reality. A statistician named George E. P. Box said “all models are wrong, but some are useful.” Objective facts are great, but they’re not actually what I’m interested in most of the time. And what I’m planning to explore in these posts reflects my truth-seeking priorities. I’m hoping to build and test and measure models to see if I can hold up those models against reality and see which ones are close enough to reality to be used and acted on.