I fortunately happened upon the Total Drek blog today, and in particular this wonderful entry combining three of my favorite things: xkcd, philosophy of social science/methodology and sociology. The post is quoted at length below with a brief discussion, but mostly you should just go read the original.
Total Drek: K.I.S.S. my ass.:
Nonetheless, the thing about K.I.S.S.[Keep It Simple Stupid] and parsimony is that you have to be very careful that you don’t get so focused on the “simple” that you forget the “solution.” In engineering or in science a solution is meant to deal with some particular problem and, while simple is good, a straightforward approach that doesn’t fix the problem is not preferred over a somewhat more complex approach that does fix the problem. K.I.S.S. and parsimony are really only good standards for judging between solutions that are equally good at dealing with problems or, at the very least, very close to equally good. I bring this up because people very frequently take a preference for simplicity to extreme lengths, and often end up paying a price for it.
To understand what I mean, consider this comic from the excellent xkcd:
Humorous as this may be, it depicts a process that is familiar to any student of social psychology and, particularly, someone conversant with Status Characteristics Theory. This theory, for those who don’t know, deals with how certain traits increase or decrease our rank within social groups. A characteristic (e.g. skin color, sex, age, training, habit, etc.) that influences one’s rank is said to confer status and, so, is a status characteristic.
Because femaleness is thought to be less valuable than maleness, the observation that a specific female is bad at a thing reinforces the original belief that femaleness is bad. Because one female is bad at math all females are now taken as being bad at math. On the other hand, because maleness is valued the observation that a specific male is bad at something is just cause to conclude that a specific male is bad at something. It’s a version of that old favorite game, “Heads I win, tails you lose,” and it helps to explain why status orders are so gosh darned resistant to change.
And this, unfortunately, has a lot to do with K.I.S.S. Diffuse status characteristics are convenient ways to make guesses about the competence of others but they provide no certainty. So, if you encounter someone who has the low value of a diffuse status characteristic, and who is poor at some task, it’s easy to explain the one with reference to the other. It is, in effect, a nice simple explanation. But, as we’ve figured out by now, simple is sometimes just another word for wrong. It’s good to keep things simple, but not so much so that our solutions don’t work anymore.
Perhaps in the end the most important lesson of all is that our judgments of other people may simply be wrong.
I really enjoyed the post as a whole. I’m not someone conversant with Status Characteristics Theory (my background in social psychology being sadly non-existent), but the explanation via xkcd here rings very true, and is a nice way of explaining the subtleties of the comic (without entirely killing the joke). I wonder how many other similar situations we can find, both in everyday sociological reasoning, but also in more scholarly work, of K.I.S.S. leading us wildly astray.
For example, I had the opportunity to attend an excellent lecture by historian of science Paul Erickson (currently a postdoc at Michigan, but bound for Wesleyan). The talk was on Erickson’s new work on the history of the concept of the population, and in particular, how asocial models of population proliferate in both studies of animals and humans, and social models tend to be ignored. One of the best examples from the talk was a story about one researcher asking a group of biologists studying rabbits if some difference in disease rates could have been caused by a difference in behavior between the rabbits in the different locations. The researchers hadn’t bothered to check for that, they said. Behavioral differences weren’t part of the model of population.
After the lecture, an eminent social scientist of a mathematical bent stood up to give a defense of the ‘simple models’ Erickson had somewhat derided. The scientist argued that simple models were essential for determining what data was relevant (missing some of the story Erickson had relayed, it seems), for teaching undergraduates (see every econ 101 class, for example), and relaying results to reporters. The last two struck me as particularly pernicious reasons to value parsimony – the scientist was arguing, basically, that we need simple models (simple to us, but somewhat complicated to the uninitiated) to make claims of knowledge to the masses, even though we know these models to be somewhere between imprecise and wildly wrong. The first reason is messier – a reading of Kuhn, for example, suggests that you can’t have observation that is not theory-laden. There is no naive collection of data possible. Alright, well and good, but can we at least not pretend to know more than we do? It makes the whole endeavor (i.e. social science) seem even more foolish than it is.
To get back to Drek’s story, parismony is good – but only when we have reasonable ways to evaluate whether or not a more parsimonious model still ‘solves’ a problem. Parsimony for parsimony’s sake – or for convincing the world you know more than you do – is painful and problematic.