It’s a commonly made claim: science depends on making metaphysical assumptions. Here the claim is being made by Kevin Mitchell, a neuroscientist and author of the book Innate: How the Wiring of Our Brains Shapes Who We Are (which I recommend).
Dawkins’s writing style does seem to divide opinion, though personally I liked the piece and consider Dawkins to be more astute on the nature of science than he is given credit for. Mitchell’s central criticism is that Dawkins fails to recognise that science must rest on metaphysics:
This is a “foundational” view of science, supposing that we build knowledge from the bottom up, and that the foundation consists of axioms of logic, mathematics and metaphysics. This idea runs into a huge problem: how do we validate metaphysics? Everyone has given up on the quest for a priori truths. And if we can’t validate metaphysics, then how can we argue that science arrives at truth if it rests on unwarranted assumptions?
I regard the foundational view of epistemology as wrong. Instead, we should adopt the view that philosophers of science call the Quinean Web view, also described, prior to Quine, by the analogy of Neurath’s raft.
This says that science arrives at a model of the world that consists of an intertwined web of ideas. We validate and improve the web by comparing the resulting world model to reality. But none of the individual ideas is “foundational” in the sense that it is axiomatic and not open to testing or challenge. Any and all ideas within the web can be assessed, and either accepted or rejected, by evaluating how well they help the overall web to model reality.
Crucially, this testing and verification extends to all parts of the web, to the logic, the mathematics, and to any “metaphysical assumptions” that you consider science to be making.
In Neurath’s raft analogy, any part of the raft can be replaced while standing on the other parts of the raft. Of course one cannot replace everything simultaneously (you’d have nowhere to stand) but one can assess what each and every part contributes to the ensemble by swapping it in and out.
We know that science arrives at sufficiently-true models of reality because it works. That is, if we use science to engineer a bridge across a gorge, then we can indeed drive a truck across it. And when aerospace engineers use science to design and build aircraft, the aircraft do fly. When we use science to design iPhones, the iPhones work. When we use science to predict eclipses, the predictions come true. When scientists develop vaccines, the vaccines work and greatly reduce disease.
The aeroplane would only fly, the bridge would only hold, if we had a sufficiently good understanding of aerodynamics, and of the nature of forces, and of the stresses and strains on materials — and a vast number of other ideas within science and engineering (where “engineering” is the application of science to real-world problems). Everyone accepts that engineering is validated and verified by the end-product working as designed.
But, just as much, the fact that the aeroplane flies also validates the mathematics used to make the calculations, and the logic underpinning all the reasoning used throughout. If “2 + 2 = 4” or modus ponens were not true (meaning, were not good models of how the world actually is) then the science and engineering that employ such tools would fail.
And the same applies to any “metaphysical assumptions”. Either: (1) such metaphysical assumptions make no difference to the predicted times of eclipses, or to the load bearing of the bridge, or to the function of iPhones, or the calculations of quantum mechanics — in which case science does therefore not depend on assuming them, or: (2) such metaphysical assumptions do make a real-world difference to the verifiable predictions of science — in which case these “assumptions” are tested and verified along with the other ideas in the overall web whenever we test scientific models against reality.
And that latter would mean that the “metaphysical assumptions” are not metaphysical at all, they are themselves scientific ideas that model reality, and are just as testable as all the other parts of scientific models.
The Quinean web understanding of science tells us that there is no good reason to make any fundamental epistemological distinction between the “scientific” ideas within the web and the “mathematical”, “logical”, and “philosophical” ideas. Such labels might be useful for different areas of knowlege (in the same way that labels such as “biology” and “chemistry” are useful), but there is not a big epistemological divide between them.
At this point, someone might be tempted to respond: but doesn’t this analysis all depend on the Quinean-web view of science being true, and surely that’s a philosophical claim, so how do you verify that?
But, again, ideas about the nature of science and how science works — along with any other relevant issue of philosophy — is just another set of ideas that is intertwined with the rest of the web of ideas with which science models the world. And the Quinean-web view is verified in the same way as any other scientific idea: it amounts to the best explanation that we have.
This is not a foundational view of epistemology, it’s a boot-strapping account of epistemology, and it’s the only one that properly explains why science is so successful in modelling the world.
If you really think that: (1) science rests on metaphysical assumptions; and that (2) these assumptions cannot be verified and so are likely wrong; and that (3) if they are wrong, then scientific ideas will not be true, then what’s your explanation for the fact that astronomers’ predictions of eclipse times do come true, that bridges carry traffic, and that iPhones work?