I find this to be an important realization with regards to atheists and the search for God.
Start with the concession the empiricist wants: scientific inference to the physical world has a discipline that pure philosophical interpretation lacks. It answers to intersubjective measurement, prediction, and the possibility of being wrong in a way that shows up publicly. When you infer the neutrino's mass or the age of a rock, the world pushes back through instruments, and the pushback is corroborable by anyone with the apparatus. That's real epistemic force, and it's why physical science is the closest thing we have to knowledge that compounds.
But the empiricist who stops there and says "only that counts as knowledge" has already sawed off the branch. That very claim — only empirically testable propositions yield knowledge — is not itself empirically testable. It's a philosophical interpretation of the evidence about which methods have worked. So the strict position is self-undermining; you cannot draw the boundary of knowledge without standing outside it and reasoning philosophically. Logic and mathematics are the cleaner case: we know theorems, and we know them more securely than any physical fact, yet not one was established by measurement. That is knowledge from valid, sound inference with no instrument in sight.
So the honest question isn't whether philosophical interpretation can ground knowledge — it plainly does, or we know almost nothing — but under what conditions it does. And here's the real distinction, which cuts across the science/philosophy line rather than along it. Reliable knowledge, in either domain, requires premises that are genuinely load-bearing and an inference that genuinely transmits their warrant. Science's advantage is that its premises are constantly re-anchored to observation, so bad premises get caught. Philosophy's vulnerability is that a valid argument from a plausible-seeming premise feels like knowledge even when the premise is quietly doing illegitimate work. Validity and even soundness aren't self-certifying — you can have a valid argument, believe the premises true, and still be wrong because a premise smuggled in more than it earned.
That's the crux for your kind of project, where moral or metaphysical facts are treated as discoverable: the inference can be impeccable and the conclusion still fail, not because philosophical interpretation can't yield knowledge, but because the premises weren't as secure as their plausibility suggested. So the discipline philosophy has to impose on itself, lacking instruments, is ruthless pressure on premises — where did this one come from, what would show it false, is it carrying weight it hasn't earned. When it does that, it produces knowledge. When it doesn't, it produces valid-looking confidence, which is more dangerous than obvious error precisely because the logic checks out.
The search for God is a metaphysical pursuit. It requires thinking outside the box as much as anything inside the box of knowledge. I say this as an atheist.
