Political Science is bullshit, Part III: What we should be doing

 In Part II of "hey, kids, don't waste your time in my classes," I went through a minimalist syllabus of core Political Science books, focusing (of course) on my primary areas of interest, in order to see if I could get away with books that are still decades old.  Yup.  What lesson do we take from this?  Not much of value is happening in Political Science.  Stuff is being written, but we are mainly nibbling around the edges, and at best making minor improvements to existing major models.  There is an "at worst," and I'll get to that in a subsequent entry in the series, but for now, let's just keep in mind that "at best."  Basically, we're stuck.  We're in a rut.  We're not going anywhere interesting.  We are adding nothing of real value to our collective knowledge.  Our efforts are being wasted.

So what should we be doing?  At this point, we return again to Thomas Kuhn's The Structure of Scientific Revolutions.  A paradigm can be used within a scientific endeavor to generate hypotheses, but eventually, a paradigm reaches the limit of its usefulness.  One of the ways it reaches the limits of its usefulness is when you begin to encounter anomalies.  Here is the trouble.  There are puzzles, and there are anomalies.  A puzzle is an empirical mystery that can be explained within your existing paradigm.  An anomaly is an empirical mystery that cannot be explained within your existing paradigm.  How can you tell the difference?  You cannot, until someone proposes a new paradigm.

And here's the thing about our empirical tests.  We use statistical models in which we accept some degree of random error, our inability to measure every relevant variable, and plenty of other limitations such that we never have a perfectly predictive model.  Is remaining unexplained variance explainable with a superior model which is fundamentally different, or is the best we can do just chipping away incrementally, and explaining another tiny proportion of the variation in our dependent variable, one new APSR paper at a time?  Unknowable without the presentation of a new paradigm.  Some day, any subfield may see one, but as of right now, today, that borders on a non-falsifiable claim.

The problem is that the difference is the difference between having reached the fuzzy border of what is knowable with an existing paradigm, and having reached the fuzzy border of what is knowable with any paradigm.  That difference matters.

Let's make an analogy to quantum mechanics.  Quantum mechanical models explain the behavior of subatomic particles with probabilistic equations.  Can we have a deterministic model?  String theory is an attempt, but what if subatomic particles really do behave probabilistically?  If so, then the process of improving our models is the process of understanding the particles and the equations within quantum mechanics.  You'll never do better than a probabilistic statement regarding one particle because according to the laws of physics, you cannot.

If existing Political Science models within some topic top out at 85% confidence, could I make a dramatic improvement with a different paradigm, or is the best I can do an incremental improvement by operating within an existing paradigm?  That latter approach is something, but arguably not that important, in most circumstances.  I'm reaching the limits of what is knowable.

Perhaps the easier question is the question of what we are not studying and should.  There was a conspicuous absence on the syllabus I provided last week:  How Democracies Die, by Levitsky & Ziblatt (2018).  I do think that this is an important book, yet it is important less for its explanatory power than for describing questions and problems that are not fully explained.

Briefly, for anyone who has forgotten (despite the fact that I bash people over the head with this book regularly), Levitsky & Ziblatt describe the process of democratic backsliding, wherein countries with democratic processes and institutions do not fall to military coups, nor anything so dramatic, but rather, have their institutions and guardrails eroded over time such that the edifices of democracy remain as empty symbolism rather than fundamental guiding forces.  Hungary is a canonical example, having democratized after the fall of communism, only to see its democratic institutions eroded by Orban, leaving a system that has the veneer of democracy without any of the operative processes.

Levitsky and Ziblatt observed several cases and noted some similarities, with observations about Trump and distressing parallels to the U.S.  Yet with limited data on a relatively new phenomenon, it was a descriptive book rather than a theoretical book.  Levitsky and Ziblatt were guided by some underlying theories about democracy and pluralism, but in many ways, the importance of the book was to direct Political Science to the problems, both as normative problems and as conceptual problems.

The term, "populism," has been stretched beyond any real meaning, so I try to avoid it, yet the far-right and anti-democratic movements across various countries, now including Italy, introduce a real set of questions for which Political Science does not have an answer, and perhaps does not yet even have the tools to answer.

We return, then, to Kuhn.  Galileo, Newton, and most of the other physicists whose names you know did not merely make empirical observations.  In order to make the conceptual and mathematical advances they made, they had to make their own improvements to existing tools such that the tools would have the necessary precision.  Galileo and Newton, in particular, are important not merely for astronomy and the equations that explain gravity, but for optics because they needed to design the optics that would allow them to make the measurements necessary to get the data necessary.  In order to gather the necessary data, you need the tools, without which you are lost.

Political Science does not have the tools to study a process that is happening, right now.  We have the tools to study past events.  The old joke is that social science isn't about predicting the future, it is about predicting the past.  We look at past data and construct models to see which variables have the most explanatory power for which other variables, in the past.  The future?  Ask your magic 8-ball.

Yet in order to do even that, we need to observe the dependent variable.  The outcome.  The endpoint.  The past.  We have terrible tools for studying ongoing processes.

What do we have?  Well, we have game theory, which is my cue to defend game theory as it is my primary methodology, but game theory sucks, except Thomas Schelling.

What should we be doing, then?  I present democratic backsliding as an example of a big problem that we do not truly have the tools to study.  Yet here is the problem.  The big questions are the questions that we cannot answer, oftentimes because we lack the tools to study them.  Political Science is so frequently engaged in the famed drunkard's search.  For those who don't know the joke, it goes like this.  A man walks down the street and sees a drunkard, on his hands and knees, leaning against a lamppost.  The man asks, "hey, buddy, can I call you a cab, or something?"

The drunkard responds, "no, I'm just lookin' for ma' keys."

"Oh, did you lose them around here?"

"Nah, I'm just lookin' 'cuz this is where the light is."

There are a few lessons here.  Lesson The First:  Don't drink and drive.  Lesson The Second:  Don't drink to excess in the first place.  Lesson The Third:  If you want to find those keys, you can either wait until morning, or get yourself a flashlight.

I'm currently fond of the Lumintop FW3A.

Hey, and opportunity for Richard Thompson!  Richard & Linda, with a live performance of the title cut from I Want To See The Bright Lights Tonight.