Ninth in a series on the edges of science
I have taken some months of hiatus in writing these essays on science—not because I ran out of topics, but simply due to personal issues.
There is no dearth of topics about the limitations of science. I’m very glad we have and use science. It has made our current level of civilization possible. Yet I do not wish to fool myself that science has all the answers.
Science uses models.
No. Not just “uses.” Without models, science hardly exists. The entire concept of science is to discover general knowledge that can be reapplied. That “general knowledge” itself is a model.
What is the nature of a model? A model is an abstraction of the thing being modeled. Science is an abstraction of reality.
In the early days of the scientific revolution, science models were mathematical or philosophical in nature. Through observation, for instance, Copernicus came up with mathematical descriptions of orbital mechanics. Newton’s famous studies of the laws of motion were mathematical. He even developed calculus as a higher form of math to deal with his equations. For over two centuries, mathematical models ruled science.
Today, the preponderance of modeling is done with software simulation. (Often based in underlying mathematical models.) We write code to describe heat transfer through conduction, convection, and radiation, then extend that code to describe the world climate. Relying on the accuracy of our models, we then use our simulations to project climate change for the future.
Models are exceedingly useful. They allow us to observe and understand things we cannot see because they are too small, too large, too complex, too fast, too slow, or too anything—or even things that do not yet exist. Models lie at the core of science.
However, as George E.P. Box famously wrote, “All models are wrong—but some are useful.”
His statement is self-evidently true, because any model is, by its very nature, not the thing being modeled. Hurricane prediction models are not hurricanes. A heat flow model of a jet engine is not the engine itself. A mathematical description of a black hole is not the black hole. The model is always a simplification of reality.
If the model is not the thing itself, then the immediate question arises: In what ways is the model different than the thing?
Models may differ from the thing modeled in the details of their behavior. While a model may be useful in general, its specifics are not. A traffic flow simulation might match the statistical parameters of an actual highway, but that is no guarantee that the individual vehicles in the model follow the exact same paths as real vehicles.
Models may differ in accuracy. The model might approximate real behavior, but individual model runs can vary wildly. Hurricane models, for instance, are quite accurate for landfall predictions two hours away—but are only general approximations for the storm position three days away.
Models may differ in applicability. A model tailored for one application might be completely useless for a different application. Cixin Liu’s novel The Three-Body Problem plays on the different climate dynamics of a world in a close binary star system. In the book, Earth engineers simply cannot understand what happens in that world because their models are based on the Sol system.
Responsible engineers and scientists validate their models by comparing model results against the reality of the thing modeled. Unfortunately, the very next step after validation is to use the model for conditions beyond the point of validation. That’s why the model was created in the first place.
So, what do these considerations say about science? If science is based on models of reality, and if models are always wrong, then can we rely on science at all? Yes, of course we can. Our very civilization depends on it.
Yet as with every aspect of science that I’ve covered in this blog series, we must always treat science with healthy skepticism. Blind belief is the realm of religion. Good science always questions itself.
So, in my science fiction, I enjoy looking for the holes in known science. Those areas in which our assumptions and beliefs may be flat wrong. There is room for some great tales in such considerations.
Doc Honour
June 2024
Doc Honour Books 