The more I read on the history of science, the more I notice a common pattern: people really want to know what is real or not.
For example, Newton introduces and reframes concepts like forces and action at a distance; immediately and for centuries every student and researcher of mechanics struggled with the true meaning of force, and whether forces are real, and what about action at a distance, that feels fake, that makes no sense, and what is even a point mass...
But getting this right is important, no? We need to figure out what is real, and what it means exactly, otherwise, what can we do?
Let's look for inspiration at the day-to-day equivalents of these deep scientific questions: is a hot dog actually a sandwich? Is a calzone truly a pizza? is a cubic croissant really a croissant?
Well, has anyone ever made better food by deeply pondering these questions? Does figuring out whether a hot dog is a sandwich helps you make better hot dogs?
Of course not. We recognize that these are mostly shitposting questions, rambling debates we like to have with our friends and coworkers just for fun. It's not deep, there's nothing that really depends on it, it just doesn't matter.
Similarly, most questions about the reality of the concepts of our scientific theories are basically worthless. Whether electrons or forces or valence or inflation truly, really, actually exist is just not a useful question. Why? Because these questions don't help us accomplish anything practical. They only change how we feel about the theories and the concept in question: whether they feel real to us, whether we're comfortable using them, whether we like them. But they don't change the maths, they don't change whether the theory works or not in a given setting.
Does that mean anything goes, and we shouldn't care about what theories mean because everything is relative?
Of course not. Gravity still pulls you down, whatever you think about it. Which is why the keyword to extract from the previous paragraphs is "use", and a use assumes a goal. What you are trying to accomplish is really the main way of judging whether you have a useful and good model/theory/approach.
If you want to predict the movement of planets, Newtonian mechanics with point masses mostly works. That means it obviously captures some key feature of the domain (celestial movements), otherwise it would not be able to predict it as well as it does.
But is it true? Well, if you start having many different bodies, using more advanced formulations of classical mechanics (like the Lagrangian one) works much more reliably. And if you really, really want precision, or are working with insanely fast and/or big objects, General Relativity will yield much better results. And if you want to do material science, you find yourself far better served with continuum mechanics than point masses.
We can discuss the reality of these different approaches, but that is not what helps engineers and scientists do their job. Instead, the sort of questions they have to ask is: "What is the minimal amount of details and complexity I need to solve my problem?"In this grounded and practical frame, a model sucks not because you feel a certain way about it, but either because it's not able to do what you want (not powerful enough) or it requires way to much work/complexity/ computation to do what you want, making it impractical (too complex).
From this perspective, General Relativity is less adapted than Lagrangian Mechanics for aerospace engineering not because it is more or less real and true, but because General Relativity brings much more complexity, for no practically useful power in this situation.
When you spend time worrying about the reality of your model, you're replacing the essential question from the last paragraphs with a fake one that, however you answer it, will never help you, and will most likely hurt you. If you decide that the concepts are really real, then you will be resistant to changing and updating them when it’s needed; if you decide that they are not real, you will underuse the model even in the range where you know empirically and practically that it works and is the efficient solution.
The moment we stop caring about this feeling of reality, and focus instead on what we can use our models for, we open the door to many fascinating epistemological questions that I will surely go back to in future posts.
Notably, when is it better to use a phenomenological model (a compression of the behavior of the system that doesn't look for deep causes, like the ideal gas law or macroeconomics) vs a mechanistic one (a model that regenerates the behavior through more fundamental and lower level entities, like statistical mechanics and micro economics); what are the different ways a model can fail to be usable, and what are the approaches to address these issues; how much is the change of goal likely to change the right model to use?
These, not empty debates about reality, are what makes epistemology worth studying. And these are what lead to better science, better engineering, better medicine.
And better hot dogs.