Rescuing logical empiricism

My philosophical journey has been a whirlwind, not far removed from my quantum meanderings. For most of the last twenty years I've been a pragmatist, starting with William James and eventually converting to Peirce.

Along the way I've dabbled quite a bit in traditional empiricism. Rationalism never really took hold with me; I've always felt we need to stay grounded in the world as it is, and that reality is whatever keeps pushing back no matter what we'd prefer to believe.

I need to admit, though, that I've burned more than my fair share of credits arguing with Claude about logical empiricism. I've always seen a lot of value in the Vienna Circle's work on meaning and truth, and in its ties to Ernst Mach, Ludwig Boltzmann, and the early architects of quantum mechanics. The more you dig, the more you realize logical empiricism and early quantum mechanics grew up together.

Popper usually gets credit for the finishing move on this school of thought (he was happy to claim it himself), and I do appreciate his demolition of historicism. But I think a lot of logical empiricism can still be rescued.

The standard attack goes something like this:

1. Logical empiricism holds that verification is the test of whether a statement means anything at all. If a claim can't, in principle, be confirmed by observation, it isn't false; it's empty.
2. Popper (and, acccckkkttttually, Hume and Mill long before him) made the black swan case. For centuries, "all swans are white" looked like a safe claim.
3. Then Europeans reached Australia and found black swans. Verifying a universal claim would have required something impossible: observing every swan that has ever existed or ever will.
4. Popper proposed falsification instead. If I make a claim, can it be proven wrong? One black swan is enough to torpedo the claim that "all swans are white."

So yeah, I'd also call that a finishing move.

But verification falls apart for universal claims, and falsification falls apart for existential ones. Say you claim "there is a graviton." To falsify that, I'd have to scour the entire universe and come up empty. To verify it, you just need to find one. Popper knew this and bit the bullet: in his framework, pure existence claims sit outside of science. That's an odd place to park a particle physicists have spent decades hunting. Weird, right?

And in practice, the search looks the same under either criterion. Whether you call yourself a verificationist or a falsificationist, you go looking, and when you find one, you assess.

My point is that we need a modified criterion, and Popper's finishing move wasn't the finishing move academia treated it as. Most likely the philosophical community, as it's wont to do, got tired of navel-gazing inside a dominant school and wanted something different. Popper offered a way out.

That's not far from what happened to pragmatism in American universities. Once fresh, it exhausted its untrodden ground, and philosophers wanted something more. Enter analytic philosophy, carried in large part by Vienna Circle émigrés like Carnap, Reichenbach, and Hempel, who shaped American departments from the 1940s through the 1970s. Then Rorty, Putnam, Haack, and others dusted off pragmatism to see what they could modernize. Rinse and repeat.

I attempted to make the data connection with Claude, asking if silent failures in ELT jobs (ie. when the job says it succeeds, but only because the source system failed to update all day) represented verification or falsification. The response:

Both, which is why it fails silently, and it's a cleaner illustration of your graviton argument than the graviton.

Your pipeline's success check is falsificationist. The implicit claim is "the job worked," and the test is "did anything throw an error?" No exception, no falsification, so the claim stands. But zero records isn't an error. Nothing broke; nothing arrived. The test was only ever looking for evidence against.

The claim you actually care about is "today's data exists." That's an existential claim, the same shape as "there is a graviton." And as you argue in the draft, falsification is blind to existential claims. No amount of error-checking will ever surface the absence of something. You need a verificationist check: find one new record with today's load date, and the claim is verified. Find none, and you know.

Well, whatever, I liked my graviton argument. I suppose the critical point is that we need to be looking at existential claims, rather than universal ones. When we work in our jobs, we’re usually asked for constrained intelligence. When we receive universal questions, we’re trained to give initial data with substantial caveats. That’s the responsible thing to do for those of us in the data field.

So: can we dust off the Vienna Circle? I’ve had Carnap and Ayer on my shelf for a long time. Might be a good opportunity to dust it off.