A (The!) Way Forward for the Social Sciences

In chemistry, at least, you get to just make things up.

Science! So often misunderstood as being cold and complicated. But actually such a human and creative enterprise. So argued Dr. Roald Hoffman in his excellent keynote at the Alexander von Humboldt Foundation’s 2014 Annual Meeting. Indeed, he argued, the technical jargony linguistic style of the academic article, invented to exclude natural philosophers such as Goethe from the field, now serves only to confuse and obscure. And chemistry itself may now move to far more creative heights, instead of simply realistically representing objects found in nature, it may start creating wholly abstract new chemical structures. Structures that may come in use in a new field called combinatorial chemistry.

But if the natural sciences are now comfortable moving away from jargony language and the realistic representation of objects “out there” in the “real world,” this confidence has not been extended to the social sciences. Indeed in many a good back and forth on the philosophy of science with Daniella Perry at UCLA it has become increasingly clear to me that the natural sciences are confident enough to be modern in the sense that they are up to date with the insights of 20th century philosophy and art, the social sciences, at least in the Anglo world, are still struggling with being modern in the old sense, that is they still aren’t over Rene Descartes (that’s NOT good!), and still struggle to find the “right” method to “truth.” That means that we are much more likely to evaluate of our models the old school way, in terms of whether they are correct and realistic, as opposed seeing them as new school, as pragmatic and abstracted representations of our objects of interest.

Ok, that was a mouthful and a headful. I apologize to and thank the intrepid reader who has had the patience to get this far. What I’m trying to say is that social science, at least in much of the English speaking world, isn’t comfortable in its own skin. There’s a reason that the economists rose to power in the last century, and have maintained this power despite the fact they are often wrong. They do a lot of math, and so what they produce, even though it is based on bad and misunderstood 19th century energy physics, and even though it is often unrealistic and wrong, the rigor and complexity give a scientific veneer and “heft” to the exercise, enough so that an economic report has heft, even when its assumptions are entirely implausible.

And indeed, what alternative do we have? Even if the language of economics is limited to what we can say with math and empirical observation there is a certain, not insignificant value, to making everyone, not least the politicians, defer to some level of formalism, as imperfect and awkward as it may be. After all, the alternative seems to be that we only base our policy decisions on raw ideology or political interest, and surely that would be worse. If jargon in the natural sciences succeeded in excluding the natural philosophers, then jargon in the social sciences can at least be a barrier to naked and naive political interests.

I’ve struggled with this issue for many years, and I’m happy to say that there is an alternative, and that the alternative does not require overturning all of economic science. We don’t have to go back to the thrilling but ultimately irresolvable debates of the political economists, or invent some wholly new way of studying society. All that is required is a small sidestep to recognize a strain close to the mainstream of economics in that three of its practitioners have gotten the “Nobel” in economics. And it is a strain which already has a community of practice, it just has not received the public recognition that neo-classical economics has achieved.

The strain is New Institutional Economics and two of the Nobel Laureates were in fact political scientists, Herbert Simon and Elinor Ostrom. And Herbert Simon gives us a whole rack of useful concepts that are more realistic than those used by neo-classical economics. Perfectly rational decision makers become boundedly-rational ones who do their best based on the information available to them. And the unrealistic cognitive demands of “maximizing individual utility” are done away with in favor of “satisficying,” that is applying rough and ready rules of thumb to get to results that are good enough to work.

Ostrom extends these concepts into game theory and then gives us a formal grammar to describe real world institutions based on this framework. The upshot being that we now have formal language to describe and analyze the impacts of the wide variety of institutions that we encounter in the real world as opposed to the rather limited palate given to us by neo-classical economics (private versus public).

Once again, who cares? At the end of the day don’t we end up with a bunch of squabbles between experts who know the niceties of various very complicated theoretical frameworks, but who never come up with a definite “answer” of what to do. Don’t we all still dream of Truman’s one handed economist? And aren’t the mathematical models still useful in that at least they come out with a number, or a range of numbers, no matter how far out the underlying assumptions would be? At least there is something quantitative and thus something seemingly concrete to react to.

Up to a few months ago I would hee and haw around this point. I agree, there is a certain value in having a number, even if we know it is wrong, provided we understand why it is wrong and in what direction and by about how much. A modeling exercise can in itself be an exercise in rigor that helps us clarify our thinking, and forces us to recognize our limitations, no matter how fancy our degrees and pedigrees may otherwise sound.

But in the context of trying to formalize my own work, I stumbled across something novel, and very powerful, which will have a revolutionary effect on how we do social science. Enter the computer!

Ok, maybe that doesn’t sound all that new or exciting. Well, ok, but think of how long math had been around before we started to incorporate it in any kind of systematic and widespread way into our social science modeling. Up through until close to the end of the 19th century natural language was still the way one tried to represent and model the social world, and in the 20th century economics was the social science that fully embraced mathematical modeling, and even it hit a major hiccup with the Great Depression. And the other social sciences have slowly adapted their concepts and methods to join in the math party, for better or worse.

But let’s admit it, all that math tends to be based on what people used to be able to do with a pen and paper. Our neo-classical economic concepts suspiciously lead us to problems that can be solved with calculus and differential equations. And all our equations tend to be linear. Why, because non-linear math makes our limited cognitive capacities hurt. So much better to make assumptions about consumers and producers having perfect rationality and perfect information, then make the math overly complicated.

And we can place the language used by computers in this continuum as well. It may be thought of as the “third symbol system” and in the long and halting progression of social science, one that will be increasingly used, even if it will never replace simple mathematical models, just as those models never replaced natural language. The advantage of computers is that they can handle such a high level of mathematical complexity, that in effect they can model our logical concepts.

And here is the kicker! Just last year Dr. Amineh Ghorbani published her dissertation on an agent based “meta-model” based on Elinor Ostrom’s framework. Ok, maybe that doesn’t make you jump for joy. Let me break it down a bit further.

The significance is that the computer scienc-y part of using computer has been moved almost fully into the background. The basic concepts on which the model is built are simple, basic, social scientific concepts. And the software has been adapted to the concepts instead of, as often happened with our use of mathematical models, our concepts being adapted to what we can model. The upshot is that the social scientist now has much larger canvass and palate to work with, and all that added complexity from adding a thing here, or a structure there, or a rule there, is now handled by the computer.

Do we worry about now having a black box? Well it’s true, we don’t know exactly what happens inside the computer. We know the structure, and inputs and the outputs of the model. But here we can make use of statistics, graphical representation, and our own judgement and intuition to interpret the results. And just as computers are really good at math, people are good at interpretation. We’re so good at interpretation that we can overinterpret, and that’s why having a model to structure our thinking was a good idea in the first place.

The MAIA model had some good reception and momentum. Six Masters theses have been completed using it, the University of Surrey and Oxford’s Environmental Change Institute have both received training, and it remained one of the top 20 most viewed articles on JASSS from publication until June 2014. I’m doing my best to see that it is adopted at the Ecologic Institute.

And maybe someday, when we social scientists are again confident enough, because we don’t have to bend our concepts to the math, well then maybe we, like Dr. Hoffman, can admit that there is a certain art to the science.


Leave a Reply

Your email address will not be published. Required fields are marked *