Book Post: The Structure of Scientific Revolutions
After my grandfather mentioned it a few times, I finally checked out a copy of this essay by Thomas Kuhn, which is over 100 pages deconstructing the practice of science as it is actually performed. My preferred view of science is that it is the thing where you decide what models of the world are correct by making predictions using those models and doing experiements to see if your predictions are correct. This is a powerful tool for truth-finding, but it so happens that there is too much truth in the world for a single person to know it all, and so science-in-practice is a community of people doing experiments. Kuhn addresses the dynamics of this community.
Kuhn’s core idea is that most science is performed under what he calls a “paradigm,” which is a set of models which define a realm of scientific inquiry. I think that the thing he is referring to is mostly the tacit knowledge held by members of a field rather than the explicit equations those members might write down if you asked them to summarize their models. To use an example from physics, he isn’t talking about the equation F=ma; he’s more interested in the class of problem-solving methods associated with that equation including free-body diagrams, torque/angular acceleration relations, and using systems of coupled equations to connect the vertical and horizontal components of the motion of a block on an inclined plane. He is interested in how those paradigms are replaced as scientists find that old models are not up to the task of describing the universe. I think the most interesting part of the essay is not actually the mechanics of revolution, but the paradigm of scientific paradigms itself. Kuhn argues that these paradigms shape the science which is done and changes the interpretation of the results. That’s either incredibly obvious or incredibly TERRIFYING. As a person who wants to use science to figure out how the universe works, I don’t want the results of experiments to depend on the mind state of the people running the experiment. His point is compelling though.
A scientific paradigm provides a direction in which to pursue research because people pursue research which can be interpretted using the tools they have available. If you find some fossils in Africa that look kind of like a plant from South America, the theory of evolution tells you that you might want to check in the ground on whatever tectonic plate would have been between those locations in the time between when the fossil was dated to come from and now for a younger fossil of some ancestor of the South American plant that looks like it has a common ancestor of the fossil you found in Africa. The paradigm also provides an interpretation of discoveries, which is the terrifying bit for me. Shouldn’t your experimental results be empirical? This is the trickiest bit of the argument to understand. Kuhn says that scientists in different paradigms are effectively living in different worlds. In the absence of a theory of evolution, you have no reason to go looking for missing links in an ancestry chart, so what do you do with this African fossil you find? Maybe you put the plant in the same genus or whatever as the South American plant and say it was just an ancient plant that died out. Even if people somehow found the younger fossil, it’s just another plant to categorize. Without the theory of evolution, there’s nothing to explain because speciation isn’t a process.
His proposed connection between models and results is particularly compelling at the frontier of particle physics, because the actual thing you’re measuring depends on the model you use. The actual evidence you usually get in nuclear physics is usually a bunch of signals from detectors which imply that particles of a certain energy reached your detector at some time. You’re generally working backwards to figure out what happened to make that particle get to your detector because the thing you want to measure has such a short lifetime that there’s no way to detect it directly. Since you can’t detect it directly, you just have to use your model to predict what stable particles will eventually be produced and what they should do, and then you report the parameters of your model which are implied by the data you got measuring stable particles after your experiment hopefully created the thing you wanted to know about. If you don’t consistently get the same parameters when you repeat the experiment, then that’s a sign that your model is wrong. If you get signals consistent with things that your model prohibits (eg. your model prohibits negative particles in an interaction, but you consistently measure one when you run the experiment) that’s also a sign that your model is wrong. It is incredibly natural as a particle physicist to think of models as a thing which changes what actual objects you think exist in the physical world.
My issue is that Kuhn projects this property onto everything. My first instintual flinch is to say that this is fine in particle physics where things are so unstable or uncertain in the quantum mechanical sense to describe as completely as we can describe a rock. Surely we don’t have to think that way about, like, the idea that genes are carried by DNA. We can literally read or construct DNA base by base and look at the proteins which come out with an electron microscope. At some point, aren’t you just looking at reality? Can’t you just report things as they are, and not worry that the thing you’re talking about is a false construct which will make no sense in the next scientific paradigm that takes over your field? If Kuhn is right, and every macroscopic science still has to worry about this, then I need to worry about this any time I look at anything or think about anything that I claim exists. Science is just a systemetized version of the act of making sense of things; it’s not some separate magesterium whose shape has nothing to do with everyday life. Technically, your eyes are a sensor which tell your brain that photons of a certain energy exist at a certain time. Your brain interprets that data and builds a model of 3D space that has cars or couches or dogs or trees or whatever in it. Your ears are sensors which tell your brain that vibrations of a certain frequency are propagating through the air. Your brain uses these vibrations to cross-check the 3D spatial model it has created and sometimes tell you where to look so you can make it more complete. Kuhn is saying that the actual stuff that you think exists changes as you learn new ways to look at the world. When I write it like that, it seems kind of obviously true, but what are the consequences of thinking this way? As an example, my wife and I were walking up a hill the other day, and she was complaining that there were these exposed patches of mud on the hill. I looked around, saw some stripes accross the nearby grass that looked like the lines you see when you mow a lawn, and immediately guessed that someone had tried to mow the grass on the hill right after a rain, and the heavy lawnmower ripped up the wet turf as it drove sideways over the hill since the mud underneath it wasn’t stable enough to hold up that weight. Where she saw a bunch of mud in her way, I saw tire tracks that had a certain width. If I was to report my findings in a scientific way I might have reported the width of the wheels that made the tracks or the hypothesized weight of the mower based off of a table of track depths I created by driving vehicles of various weights sideways on wet hills. That would sure seem scientific. Maybe I was totally wrong though! I don’t know that someone mowed a hill right after a rain. The world is big and complicated, and sometimes the first explanation you come up with isn’t the right one.
Thinking of science that way is hard for me. I implicitly think of experimental results as an unbiased statement of truth in the world. Kuhn is trying to break that model of science. He complains that people are misled by scientific textbooks which are written as if they hold the simple truth which was the inevitable conclusion reached by observing the pile of hand-picked experimental evidence helpfully provided by the book. The process of figuring out new scientific models, which is the thing Kuhn actually wants to talk about, is messy. Even if people are finding issues with an old model, sometimes the new model doesn’t quite explain everything the old model does yet because it hasn’t had decades of people twisting their minds to fit the model and gluing weird hacks to the model to make it work. Some bizzare scientists like the new model anyway because they think the old weird hacks are ugly and the new model is prettier, so they go grab a bunch of evidence, interpret it under the new model, and glue their own prettier hacks to the new model until it works so well that new scientists use their model instead of the old one. Suddenly all of the old truths are reinterpretted in the framework of the new model! The new model is the way the world always was! And so the way the world works becomes different than it was a few decades ago.