I’ve worked with a lot of numbers people and never met any [except physicists, who don’t count] who believed that their work was doing anything except helping to inform a bigger theory.

(In double-checking this in OED, I was interested to learn that it was Linnaeus who applied the word to the grub of an insect; the underlying metaphor is that the grub is a ghost or appearance of the “imago,” the adult “true form” of the insect.)

We have such a dichotomy in medicine as well.

There’s this upstart school of thought that calls itself “evidence-based medicine”, which sounds like something pretty hard to be against (I mean, what’s the alternative, “faith-based medicine”, “intuition-based medicine”, “wild hunch-based medicine”, “sheer ego-based medicine”, “I’ve-been-a-doctor-longer-than-you-so-shut-up-based medicine”? Lord knows, we see all of these schools of thought in action.), except that they’re really talking about quantitative evidence only.

The problem is that medicine is still mostly an observational science. There’s a thin veneer of potentially quantitative evidence-based therapeutics riding on top of a base made up of unquantifiable pattern recognition. To get a science of medicine, first you have to divide off that bit of human misery that we treat from the rest of human misery (I mean, I can do something for you if you come to me with the pain of a blocked coronary artery, but I have nothing to offer for the pain of a broken heart.), then we had to subsivide our turf into its separate syndromes and diseases, and then place these disorders in their proper classification within the universe of medically-treatable human misery. Once you have your ordered list of disorders, then you carefully observe the natural history of all of these disorders. Individual providers confronted with individual patients presenting with an array of signs and symptoms then have to sort out which disorder/s, if any, are present, or whether this patient has a case of non-medical human misery. Only after we get through all of this process, which is all gestalt and pattern recognition and not quantitative at all, do we get to treatment, where quantitative methods finally are of use in distinguishing what works from what doesn’t.

Okay, but look, you say, isn’t all that preliminary stuff prior to treatment, at least for medicine in general, already done and settled? Of course, there will always be the non-quantitative problem of recognizing which disorder/s best fit the pattern of symptoms and signs presented by the individual patient, there’s always a seat-of-the pants basis to diagnosis, but you would think that at the general level, that list of disorders and their natural histories would be a settled matter, and all the action that’s left would be at that thin veneer of therapeutics where quantitative methods reign.

Sadly for those who would prefer that life be simple, that’s not how it works. Every new treatment, and even every new lab test or imaging modality, rejiggers the natural course of disease so that we have to start observing all over again to re-establish a new natural history. We start doing screening for prostate cancer because now we have a new lab, an assay for PSA, and we don’t know what to do with all of the early prostate cancer we discover because all we know is the natural history of clinically-discovered prostate cancer. There’s no guarantee whatsoever that screening-discovered prostate cancer is at all as dangerous or rapidly progressive as the clinically discovered prostate cancer we already know about, and some reason to believe the opposite. But until we’ve had a chance to observe the new entity, the idea of just assuming that it is probably slow-growing and doesn’t need to be treated as aggressively as clinically-discovered prostate cancer could obviously condemn patients to possibly preventable deaths unless it is treated aggressively. It won’t be ethical to bring out our quantitative methods and do a randomized control trial until and unless old-fashioned clinical experience accumulates which suggests that it really is a toss-up whether or not screening-discovered prostate cancer benefits from immediate aggressive treatment. By the time that study is done and reported, the folks actually treating patients with prostate cancer will have already moved on to not treating aggressively, based on non-evidence-based clinical experience.

So, by all means quantify what you can, and what you can establish by such means is of outsized importance precisely because of the rarity of such anchoring points of solidity. But quantitative methods just won’t often get you far from your base of “I’ve-seen-this-based medicine”.

One day physicists will discover[/admit] that their objects of study are as kludged-up as those of biologists, and then where will we be?

Isn’t physics almost by definition the study of objects that AREN’T like that? If it gets too hairy it becomes chemistry. (Tongue inserted only partway into cheek.)

And again, you’re not the first person to have noticed this. Assuming that one accepts that the ‘hard sciences’ are physics, chemistry and biology, it’s still bizarre that, when attempting to make their (‘soft’) science more ‘scientific’ , scholars in the social sciences almost invariably choose to model their science (or attempt to model it) on physics rather than biology (a rather more obvious choice one might think). I think it was Mary Midgely who first pointed this out.

The reason, presumably, is that the objects in physics are more amenable to abstract mathematical modelling than those of biology.

I don’t see why human social behavior should be expected to be any less complex, and any less historically conditioned (by both evolutionary and cultural history in the case of human behavior), than molecular biology, or any more successfully modeled by oversimplified quantitative models of the kind physicists employ. There’s a well-known joke about physicists that I’ve always enjoyed, in which a physicist is called in by a dairy farmer to help solve a problem with his cows’ milk production. The physicist’s report begins, “Let us model a cow as a perfect sphere…”