The Duhem-Quine Solution & ‘Explanation’

The Problem

Popper, Hume, Russell and others pointed out the logical asymmetry between verification and falsification: observing one black swan can refute the theory “all swans are white”, but no amount of observing white swans can confirm it.

(If a theory T implies consequence X, then ¬X implies ¬T. But neither X, nor any other consequence of theory T, implies T.)

The problem is: theories don’t translate neatly to logic.

Say you experimentally test a theory, and you predict a consequence X, but instead the experiment’s result is ¬X.

In logic, ¬X implies ¬T.

But in real life, the experimental apparatus might be faulty, or it might be the wrong experiment to show ¬X, or you might have made a human error and miscounted something when doing the experiment.

So you can never know if your ¬X really is ¬X.

This was noticed (and called the ‘Duhem-Quine thesis’), and people thought, “Given the logical asymmetry doesn’t work in real life, that means Popperian epistemology doesn’t work – because Popper is based on this asymmetry between falsification instead of verification, which is only valid in pure logic and not in scientific methodology”.

Popper tried to answer this in LScD (part 20). He said two things:

1. All objections to the refutation (¬X) themselves must also be falsifiable.
2. An objection works if it would “constitute real advancement in our knowledge of the world”.

The first one doesn’t actually solve the Duhem-Quine problem – it simply moves it (because the falsification.. of the objection.. of the refutation.. could itself be false – and so on; it’s an infinite regress).

The second works better, but it’s vague, and it kinda appears to be talking past Duhem-Quine.

Asymmetry Confusion

The actual solution is that Popperian epistemology is not based on the logical asymmetry. It’s based on a different asymmetry.

That asymmetry is what David Deutsch called the difference between explanation and non-explanation.

… Which may sound like a new different theory and not in the original Popper. Until you look at what it means.

An ‘explanation’ is a statement about why something is the way it is – as opposed to only a prediction of observations (aka an instrumental statement).

Popper criticises and rejects instrumentalism (the idea that there are only instrumental statements, no explanatory ones). According to Popper, any instrumental statement must be accompanied by a theory (that isn’t just another instrumental statement), because all observation is theory-laden.

Realism is also incompatible with the instrumentalist interpretation (in other words: there really is stuff out there that we can know about, not just the observations/what we see). Critical Rationalism requires realism. Critical Rationalism therefore requires explanatory knowledge (even if this was just implicit in Popper’s writings).

So, going back to Popper’s response to Duhem-Quine about objections that “constitute real advancement in our knowledge of the world”: what Popper means there, implicitly, is that some objections explain part of the world, and some objections don’t have this explanatory component.

(This, I think, is the only way to make sense of what he’s saying there. And it would already be implied by his realism, his “all observation is theory-laden”, and his anti-instrumentalism.)

The reason I said it “kinda appears to be talking past Duhem-Quine” is actually because Duhem-Quine is kinda talking past Popper. The original Duhem-Quine thesis is ‘since all observations are theory-laden, refutations are not final any more than confirmations are’. This is true, but not relevant, because Popper isn’t looking for final refutations.

Improved Duhem-Quine Problem (Still Doesn’t Work)

If you took Popperian epistemology seriously, a variant of Duhem-Quine could be:

“Why do we think a refutation works, given you can always criticise that? Why do we adopt refutations as the new ideas at all?”