Regarding my previous post about falsifiability, here are some logical and philosophical problems with falsifiability as a criterion of knowledge (science).
1) Historical problems: The first problem with falsifiability is with what scientists actually do: Falsifiability does not explain what scientists have done, or do, when they are presented with tests that falsify their theories. Scientists have shown a strong tendency to keep their theories intact (in short term that is). They do not tend to, rightfully, throw away a theory because of one test.
This happens to be true about Newton and Einstein, that decided to keep their theories even when a test showed the theory appeared to be false. And obviously, it was the test that was false, not the theory. We can clearly see that this was the right thing: Imagine if Newton had thrown away his theory just because some tests about the moon had shown unpredicted results.
The answer of an advocate of falsifiability to this historical problem is fairly simple, and I think acceptable to some extent. Firstly, the test must be repeatable, which means it’s not just one test, but one test that repeatingly keeps showing abnormal results. It’s still the same test, but here we could be fairly sure that we are not doing anything rush. Also, another answer is falsifiability is not a historical method, but an imperative methodology: It’s saying what scientists should be doing, not what they have done.
2) The practical problem of Duhem – Quine thesis: The thesis itself is simple, we cannot test any theories without testing assumptions and other theories that have made them in the first place. Take for example theory T which is made of other assumptions and theories t1, t2 and t3. When we test T, we are also testing t1, t2 and t3. This is not so much of a problem for other methodologies as it is for falsification. The question is, what has been falsified? T itself, or any of t1, t2 or t3?
Let’s logically formulate this:
T=(t1 ∧ t2 ∧ t3)
We should remember that in order to falsify T, we either need to have falsified T itself, or any other one of t1, t2 or t3. The problem here is more likely practical: How do we know which is actually falsified? Assuming that we can know all of t1, t2 and t3.
The advocate of falsifiability here will have an answer, which seems logically plausible, but in practice turns out very hard or even impossible (if we take Quine seriously): As long as the theory is not falsified, we continue testing it. When it does falsify, well there is no other way and we have to test our other sub theory-assumptions as well.
So, falsifiability as a methodology is not likely to be practical. But then rises a logical problem from Duhem-Quine thesis, which is very problematic for falsifiability as a criterion of knowledge.
3) The logical problem of Duhem-Quine thesis: The formulated theory T above still works here, let’s look at it again:
T=(t1 ∧ t2 ∧ t3)
What will happen if t1 is non-falsifiable? As we can see above, T is still falsifiable. It will be falsified if T or t2 or t3 are falsified. So, what happens to the idea of falsifiability being a criterion of knowledge (science)? If “any” of the pre-assumptions that we had in order to reach to T turn out to be non-falsifiable, it will show that science is not all falsifiable.
The most fundamentals of science turn out to be non-falsifiable: “The world is real” or maybe better “We can know objective things about the world”. If we can know that we cannot know, then we can know, therefore this statement is not falsifiable. It always has to be true, since if it is not true, it is true.
In the end, falsifiability by itself will not be enough. It has been suggested that we add things to it, or abandon it and take a whole other rout. In any case, although it may be a good way to show what is “not” scientific, it is not a good way of showing what is. And although it can work about some areas of knowledge which are somehow obviously not scientific, about those grey areas is not as useful.