Some people have pointed to certain “theories” as being extremely “robust”, which means that they can fit any observations we make in the real world. Far from being a strength of those “theories” this “robustness” is a fundamental flaw.
In formal logic there’s a kind of statement called a “tautology.” This is a statement where, no matter what the state of the various parameters of it, the statement itself works out as true. Given the statement p OR NOT(p) the statement is true regardless of whether P is true.
One fact about tautologies. They can tell you nothing about state of the variables involved. Consider: “either it is raining or it is not raining”. This statement is always true. Either liquid water is falling from clouds in the sky making one side of the “or”, and thus the entire statement, true, or liquid water is not falling from clouds in the sky, making the other side of the “or”, and thus the entire statement true. The truth of the statement tells you nothing about whether you need an umbrella or not.
To tell you anything about the world, a statement must have some possible conditions, at least in potential–whether they occur in the real world or not–where the statement would be true and others, again at least in potential, where it would be false. In that case, the truth of the statement gives you information about the world. “It is training” might be true, might not be, but if true than we know that an umbrella could be advisable. In order to tell you anything about the world, a statement has to exclude possibilities. It must also permit possibilities–a statement that is always false no matter the state of the elements in it is called a contradiction and is similarly useless in conveying information about the world.
These are simple examples and may seem trivial but the concept is extremely general. Any statement, no matter how simple, or how complex, that is always true (or always false for that matter) regardless of the state of the various elements in it is a tautology and cannot actually convey information.
This is an important concept in the physical sciences. There must be possible observational results that, if observed, would lead to the conclusion that the theory is wrong. Without that, it’s a tautology and conveys no information. It’s not right and it’s not even wrong. It’s not meaningful enough to be right or wrong. It’s just empty words.
The late Richard Feynman described this process in his famous physics lectures. How to find new laws of nature.
- Guess what the new law might be. (Generally after much observation to try to discern a pattern.)
- Calculate what must happen if your guess is true. This might be an individual result–a rock will fall when dropped–or it might be something of a statistical nature–half of the electrons emitted will be spin up and half spin down. But either way the specific result expected must be calculated.
- Compare the results of the calculation to experiment. Now, some people will object that some things you can’t do in a lab. No, you can’t. It’s rather hard to create a star and observe its evolution to test theories of stellar evolution. However, you don’t need to do that. You take the predictions of what we would observe from existing stars then look out with our instruments at the sky and see if what we see matches those predictions. Same with other things that can’t be done in a lab. Determine (calculate) what must happen in nature then go look.
- If experiment does not agree with the results of your calculations you’re wrong. When observation and theory don’t match it’s the theory that must go not the other way around. Yes, observational error can happen. You can have non-representative data where some other factor is affecting the results and it’s valid to check for that. But not fitting your theory is not reason by itself to dismiss the data. And you can’t use your theory to “correct” the data or even as a benchmark to say when to stop. The data must stand or fall on its own and the theory either fits it…or not. If you want to exclude “bad data” you’d better be able to justify it without appeal to your theory. If you want to “adjust” the data for measurement error, you’d better be able to justify that without appeal to your theory. And you’d better be prepared to show the original data, the exclusions and adjustments, and the justifications for them so that others can validate the work.
- It doesn’t matter how beautiful your theory is. It doesn’t matter how smart you are or what degrees you have. None of that matters. All that matters is whether or not it agrees with observation.
- Note, though, that you make the calculation first. You don’t go through all the data available and pick some that seems to fit your initial guess. You need to make your testable predictions first, then look to see if they hold up. If you simply grab on things that happen to agree you’re not doing science. If your calculations say that A must happen and B can’t, and you look and see both A and B happening, well, guess what. You’re wrong. All of the predictions that come out of the theory must hold for the theory to hold.
- One thing that Feynman did not say was when you’d know the theory was right. The reason for that is simply. You can’t. The most you can ever say is that it is consistent with available data. There always remains the possibility of new data invalidating the theory and having to start over at “guess.”
In the end, a theory that can never be falsified, that no possible data can ever overturn is no theory at all. It’s neither right nor wrong. It’s too meaningless even to rise to the level of “wrong.” But it might well be very useful politically.
I’m sure we can all think of examples.