In my last post, I discussed how Popper scholar Danny Frederick and Physicist David Deutsch (also a Popperian) actually use the word ‘refutation’ in distinctly different ways. To Frederick, a ‘refutation’ is any missed prediction. Because the Duhem-Quine thesis is correct, a ‘refutation’ in this sense must always be not of a single theory but of that theory plus all background knowledge. (See this post for a summary of Danny’s view.) To Deutsch, a ‘refutation’ is what happens when you finally conjecture a good rival theory and then find crucial tests that refute one of the two theories but not the other. (See this post for a summary of Deutsch’s view.) There could be a considerable distance in time between these two ‘events’.
The realization that the word ‘refutation’ has two fairly legitimate possible meanings led me to research how Popper used the term. And, as documented in the last post, it turns out Popper (while not 100% consistent) utilizes the term far more consistent with Fredrick’s interpretation.
However, I also noted in that post that Deutch’s understanding of a ‘refutation’ comes closer to how most people would understand that term because it doesn’t require you to ‘refute’ a combination of the theory plus the background knowledge. You are, in some sense, refuting the actual theory once you come up with a crucial test between two rivals.
But I also pointed out that even this isn’t quite right: because the scientific community rarely accepts a theory as refuted right away. Each skeptical scientist must be given a chance to try to refute the refutation, so to speak. Only if no one can find another way to solve the problem the crucial test represents (and moreover, if the new theory proves more productive) will eventually the theory be considered refuted (or to use the Frederick term ‘rejected’) by the community.
So the simple truth is that neither event fits well with the word ‘refutation’ as used in common parlance. Both require you to ignore what the word generally means and to re-imagine it as either applying to the theory plus the background knowledge or as no longer being a single event.
One possible argument against what I’m saying that I’d find to be a valid argument is that we might think of the Deutsch version of ‘refutation’ as not being for the full scientific community but for a single scientist. This would make sense as each scientist must, on their own, come to decide when the old theory is now ‘refuted.’ But even here, you can’t really say the Deutsch version of ‘refutation’ is quite correct either. Namely, because if we’re talking about any individual scientist, they are free to see a theory as ‘refuted’ the moment they see the first problematic experimental outcome. The reason why is that scientists generally have a hunch as to what part of the theory plus the background knowledge is the problem due to the nature of the problem itself. As Popper puts this:
There is first the layered structure of our theories… This structure allows us to distinguish between more risky or exposed parts of our theory, and other parts which we may–comparatively speaking—take for granted ain testing an exposed hypothesis. treating the rest of the theories involved in the test as more or less unproblematic. … Our scientific procedures are never based entirely on rules; gueses and hunches are always invovled…Realism and the Aim of Science, P. 187-188
I had one Twit Rat put it to me like this: A Popperian refutation/falsification comes with a guess as to what part of the system of theories (theory plus background knowledge) is the problem. The scientist can, from there, work on devising a test to test if he is correct. Let’s say he thinks the problem is with his instrument, then he will devise a test to make sure his instrument is working. If his test of his instruments fails to find a problem (i.e. corroborates the theory that his instruments are working) he may be forced to go back and consider a different part of the theoretical system to challenge.
But even if we think of a refutation in this way, it’s still defining the word ‘refutation’ differently than most people would understand the term: who defines a ‘refutation of a theory’ as ‘both a counterexample and a guess as to whether or not that counterexample applies to the theory or the background knowledge”?
There seems to be no way out of this: the word ‘refutation’ just isn’t the best term to describe what Popper had in mind any more than the word ‘verification’ is. (See this post for discussion.) The problem is that Popper co-opted the terms ‘refutation’ and ‘verification’ from their logical meanings rather than in how a scientist or layman would try to understand those terms. So both terms carry with them quite a bit of baggage that leads people to regularly misunderstand Popper’s epistemology.
When people hear that you can ‘refute a theory’ with a single counterexample, they sincerely think that means you have demonstrated that the theory is now false. They don’t have in mind that you might mean you refuted the background knowledge or that it was only a guess that it’s not the background knowledge. Nor do they have in mind that it might take decades for the scientific community to catch on. In other words, people understand the term ‘refutation’ far too much in a non-fallibilist mindset–not unlike how people often misunderstand the term ‘verification.’ This does lead to misunderstandings. Consider the example I documented here. The speaker misunderstands Popper’s epistemology for no reason other than he can’t get out into his mind that a ‘refutation’ always includes the background knowledge. This just isn’t how people understand the term. We’re practically encouraging misunderstandings.
But this then leads to an interesting idea: Can we reword Popper’s epistemology in such a way that we don’t need the misleading terms. That is to say, we use neither the word refutation nor the word verification?
I am going to attempt to do just that in the next post. I’ve made a number of decisions in how to word things that I suspect many will take exception to. Here is my translation guide:
|Popper’s Old Term||My New Term||Definition|
|Refutation||Counterexample / Anomalous Observation||A basic statement that contradicts the whole theoretical system (i.e. theory plus background knowledge.)|
|Verification||Positive Instance||A basic statement that does not contradict the predictions of a theory because it matches the predictions made by the theory.|
|Corroboration||Support||An attempt to test a theory that comes up with a positive instance. Not useful except in a crucial test with a rival theory. I.e. ‘the test supports the theory’ should be taken as identical to ‘the test corroborates the theory.’|
|Rejection / Acceptance||Popper had no equivalent to this term. But it is when an individual scientist decides that they have found done a successful crucial test between an old and new theory and found that the old theory did not survive the test. This may require repeatability to be sure there was no mistake in the test. A synonym is ‘Acceptance’ of the theory that wasn’t ‘rejected.’ Acceptance and Rejection are logically identical. Popper downplays the importance of this event.|
|Criticism||Problem||I suppose I’m not really changing this one much. To Popper, a ‘refutation’ was specifically empirical. By comparison, a criticism might be a refutation or it might be some other form of criticism that might be non-empirical and thus can apply to metaphysical theories. I will more or less use criticism and problem interchangeably. So for me, a counterexample is a problem but not all problems are counterexamples.|
|Conjecture||I’m not really changing this one. I may be more explicit by saying ‘conjectured solution to a problem’ or the like. I almost have to do this because Popper called a shorthand for his epistemology ‘conjecture and refutation’ which doesn’t sound right as “Conjectures and Counterexamples” or “Conjectures and Problems.” So the new shorthand must now be “Problems and Conjectured Solutions.” As a bonus, this fixes the temporal ordering! Because according to Popper we always start with problems, not conjectures. (Okay, granted, if you take ‘conjecture’ to mean ‘theory’ or ‘idea’ then it doesn’t matter which order you put it in.)|
The motivation here is to state Popper’s entire epistemology in this more natural language and see if this makes Popper’s epistemology easier to understand and, with some luck, easier for non-Popperians to embrace.
I note that the reason I chose the word ‘support’ instead ‘corroboration’ was precisely because I feel Popperians have alienated too many non-Popperians with claims that confirmations/support etc are impossible when most people know plain well that they are possible. While the word ‘verification’ is problematic, the word ‘support’ really isn’t. Since Popper himself called ‘corroboration’ a form of support, as quoted in this post here, Popperians should learn to not object to this word. For our purposes, it now just means we tried to test our theory by seeking a counterinstance and failed to find them. I feel this understanding of ‘support’ is more or less identical to what non-Popperians actually mean when they talk about a test supporting a theory, so I’m trying to pick words that seem more natural to them.