Oh Sacred Science, Hallowed Be Thy Name

Is economics a science? Is sociology? What about history? At times this whole debate seems to me to be little different than the argument over whether or not video games are an art form — it’s more about the social status conferred on labels like “art” and “science” than about the substance of any art or any science. But not to worry — Pascal-Emmanuel Gobry has come along to draw an utterly conventional positivist line in the sand on the subject of science, with economics comfortably on the other side.

Gobry lays down some Hard Facts on his readers:

So let me explain what science actually is. Science is the process through which we derive reliable predictive rules through controlled experimentation. That’s the science that gives us airplanes and flu vaccines and the Internet.

Gobry thus connects himself with a tradition which includes Popper and the members of the Vienna circle, and which he traces back to Francis Bacon. This tradition draws hard demarcations — it is full of methodological zealots who pass judgment on that which must be cast into the category of Not Science. In Gobry’s case he at least feels strongly that everything else in the humanities and social (non)sciences have real and valuable insights (and a good thing too, since those are his areas of greater expertise). But I’d like to look at a few problems that emerge if you take his demarcation too seriously.

For one thing, Ptolemy’s astronomy is a science by this definition, and Copernicus’ is not. I do not mean by this that Gobry’s criterion is wrong because it favors Ptolemy. In fact, I think Ptolemy gets a short shrift by people who idolize the so-called Scientific Revolution and think of everything that came before it as mere primitive superstition and irrationalism. Ptolemy actually made lasting and practical contributions.

There is no doubt that the geocentric model is factually wrong — we now have the instruments not only to discern this from where we are, but to go out into space and get the bigger picture. But Ptolemy made a huge number of systematic observations of the stars and the planets, and put together a model to predict their motions. Not only did his model reliably assist sailors navigating at sea for hundreds of years — it is in fact still used for naval navigation to this day!

Copernicus, by contrast, primarily advanced a hypothesis — the Earth actually revolved around the Sun, and not the other way around. Compared to Ptolemy, he did not “derive reliable predictive rules through controlled experimentation.” Of course, neither Ptolemy, nor Kepler or Galileo, derived their “predictive rules” through “controlled experimentation”; they did it from observation, and by reasoning within the framework of the scholarly culture they were a part of. Though Gobry thinks that Aristotle “set humanity back several centuries” via his epistemology, Galileo’s work on inertia was largely inspired by the theory of impetus which was developed hundreds of years prior by people operating within an Aristotelian framework. To say nothing of the other scientific advances that occurred during the middle ages.

Gobry’s fixation on controlled experimentation, especially randomized trials, is not uncommon among those looking for some magical explanation of why physics is so much more precise. But this is misguided. To see why, consider the following thought experiment (another useful tool for advancing understanding).

Let’s start with a hypothetical, stable probability distribution. An idealized six-sided die has a perfect 1/6 chance of any one of its sides coming up when you roll it. No matter how many times you roll it, this stays the case. If there were two million dice in the world, and you wanted to generalize this point about them, you could select 1000 of them (or whatever size for whatever level of sampling error you’re willing to tolerate) and roll them all and from that get a pretty reliable idea of the underlying probability of any one side coming up, given the stability of that probability distribution.

Now say someone adds metal to the number 5 side of the dice, and rolls them on a table that has a magnet on one part of it. The introduction of the metal side and the magnet changes the probability distribution. This introduction is what is called a non-ergodic event. If the dice system remained the same from the moment it is first observed until the last human drew breath, we would call it ergodic — -a perfectly stable and unchanging probability distribution. More contingently, we could say that dice exhibiting certain characteristics thrown on a surface also exhibiting certain characteristics are ergodic.

What’s the point of all of this?

So long as the underlying probability distributions of human behavior are stable, then Gobry is correct in asserting that randomized trials will reveal a great deal. Under those conditions, and to the extent that we can get truly random trials (and this in itself is problematic), we can approach the level of knowledge we have about the probability of dice rolls in our knowledge of the effects of fiscal policy, say.

Our problem is not merely that human systems are probably non-ergodic or occasionally experience non-ergodic events. The problem is that we can’t even really observe how ergodic it is, directly. This isn’t like adding metal and magnets to dice rolls. For all we know, because of the state of technology, the prevalence of certain ideologies or practices, or non-linearities in demographic distributions, we could be getting radically different probability distributions for human behavior on a very regular basis. And that’s just picking out some features I can imagine off the top of my head — Hayek and Knight, to say nothing of Taleb, would emphasize the parameters we don’t even realize we’re not thinking about or unable to observe. Controlled trials aren’t much help if you can’t say something significant about the ergodicity, or linearity, of the system under observation. I’m not denying our ability to say something meaningful, or make progress in our knowledge about, human social systems. But for my money, I think Deirdre McCloskey is a much more valuable guide than Pascal-Emmanuel Gobry.

I appreciate that Gobry really does believe that all of the knowledge outside of his very narrow specification is extremely valuable. I don’t know why he thinks that making that specification and trumpeting it to the world so loudly is important (wait, yes I do). The concept of ergodicity, the Copernican revolution and the system it replaced, and quantum physics, along with the laws of supply and demand, and the concept of comparative advantage, are all valuable theoretical tools which do not fit cleanly into Gobry’s straightjacket. In short, if deriving “reliable predictive rules through controlled experimentation” is all there is to science, then almost none of our valuable knowledge counts as science.