Woe the humanities ?

A few years ago I wrote a piece describing how valuable the humanities courses are for teaching critical thinking, here broadly in the sense of "getting at the real truth of a statement". At high school, you can't do that when teaching trigonometry. You can't have kids wondering if the sine of an angle is really the ratio of the opposite to hypotenuse, because it just is. Science at that level deals with hard facts. But language and literature can leap beyond that to the world of "what does this really mean ?" To be aware that a poem conveys information beyond its superficial surface content, which is often barely intelligible, is an important step in realising how emotions manipulate you. If you can disentangle Shakespeare, then understanding the effects of advertising and political slogans becomes a whole lot easier.

So I'm 100% in favour of teaching humanities. Not just because critical thinking is a valuable spin-off, but solely for their own sake. It is a fundamentally good thing to explore the irrational side of humanity. This means I'm sympathetic to the author of the embedded piece when she says :

Most of these disciplines aren’t quantifiable, scientific, or precise. They are messy and complicated. And when you try to straighten out the tangle, you may find that you lose far more than you gain... The tools of hard science have a part to play, but they are far from the whole story. Forget the qualitative, unquantifiable and irreducible elements, and you are left with so much junk.

Sometimes, there is no easy approach to studying the intricate vagaries that are the human mind and human behavior. Sometimes, we have to be okay with qualitative questions and approaches that, while reliable and valid and experimentally sound, do not lend themselves to an easy linear narrative—or a narrative that has a base in hard science or concrete math and statistics. Psychology is not a natural science. It’s a social science. And it shouldn’t try to be what it’s not.

But what I don't have an answer to is where, specifically, to draw the line. Humanities subjects have irreducible, unquantifiable aspects to them, but they also do have quantitative elements. Under what conditions, if any, should we use the quantitative to inform about the qualitative ?

You might remember my agent-based project to model talent versus luck*. I found this a very interesting way to illustrate and explore different possible scenarios. It was helpful to show how the data could be incredibly misleading and how correlation and causation could be confused in both directions. In exploring the data I thought about the concept of a meritocracy in a deeper and different way to how I would otherwise. The process had value. But at no point was I under any illusion that the models had any direct bearing on the real world : they could never be used to make predictions or inform policy; their use was more oblique than that.

* I basically abandoned the project when I reached a satisfactory point. I even wrote it up in a more academic-paper style format, but I got too bored to submit it.

So models can have value beyond making direct quantitative predictions. But is it fair to criticise those humanity models which do claim to be able to make predictions or understand specific situations ?

Witness the rise of Cliodynamics : the use of scientific methodology (nonlinear mathematics, computer simulations, large-N statistical analyses, information technologies) to illuminate historical events – and, presumably, be able to predict when future “cycles” will occur. Sure, there might be some insights gained. Economist Herbert Gintis calls the benefit analogous to an airplane’s black box: you can’t predict future plane crashes, but at least you can analyze what went wrong in the past. But when it comes to historical events—not nearly as defined or tangible or precise as a plane crash—so many things can easily prevent even that benefit from being realized.

To be of equal use, each quantitative analysis must rely on comparable data – but historical records are spotty and available proxies differ from event to event, issues that don’t plague something like a plane crash. What’s more, each conclusion, each analysis, each input and output must be justified and qualified (same root as qualitative; coincidence?) by a historian who knows—really knows—what he’s doing. But can't you just see the models taking on a life of their own, being used to make political statements and flashy headlines? It's happened before. Time and time again. And what does history do, according to the cliodynamists, if not repeat itself?

The cliodynamists, just like everyone else, will only know which cyclical predictions were accurate after the fact. Forgotten will be all of those that were totally wrong. And the analysts of myths only wait for the hits to make their point—but how many narratives that are obviously not based in reality have similar patterns? And whose reality are we dealing with, anyway? We’re not living in Isaac Asimov’s Foundation, with its psychohistorical trends and aspirations—as much as it would be easier if we were.

The author may be right, I don't know. There's a deep, unanswered and perhaps unanswerable philosophical question here : can the unquantifiable affect the quantifiable ? Is human psychology simply impossible to predict ? (Obligatory other mentions : my own abysmal history of predictions; the concept of deliberately seeking out or creating wrong predictions as an extreme form of the straw man fallacy)

I lean cautiously towards "no". My lingering suspicion is that individual humans are far too complicated to model accurately in a given situation, but that the statistical behaviour of a group (or an individual over time) is easier. You get anomalies and oddballs in every group that defies the major trend, but you can still often pick out the major trend. And the map is not the territory, but it's often good enough. You cannot quantify justice, but you can, maybe, quantify how people feel about it.

Now, only an insane person would claim that modelling the behaviour of a society is simple enough that we can expect predictive-level modelling of political issues anytime soon. But it seems to me a mistake not to try. It might be that some things literally cannot be quantified and are utterly impossible to model in any way whatsoever, but it could also be that things are just fantastically complicated and we don't have or understand all the variables just yet. See, humans are often quite easy to predict - they're often creatures of mindless habit. The difficult bit is the specifics : when they deviate from their habits, what event will be the straw that breaks the camel's back.

My take-home message would then be : humanities aren't a science, but they have scientific aspects. Try and model them but don't expect the same kind of results as in the hard sciences. If you can use them to model a scenario directly, then great, but if not, their less predictive aspects aren't necessarily less valuable.

Humanities aren't a science. Stop treating them like one.

There's a certain allure to the elegance of mathematics, the precision of the hard sciences. That much is undeniable. But does the appeal mean that quantitative approaches are always germane? Hardly-and I doubt anyone would argue the contrary.