Currently, I’m trying to find my way through the standard data analysis toolbox (the shell, git, markdown, etc). Something that has become apparent to me is that, although I have been coding now for several years, I still find there is a considerable learning curve for these tools. Even when you are a true believer, it is inevitable to ask yourself about, why should we care? Why work in Markdown or, worse, LaTeX. if most of what you are going to write can be done, allegedly “more easily” in Word? Why learn to do data analysis in R or Python, when you could be doing the same in SPSS or even Excel? Why model a game with abstract parameters and specify your result for a certain region when you could be making the same point with numbers? Who needs math if the computer is likely to automate everything for you?
To be sure, I don’t think these assertions are generally true. Many ideas can not be appreciated if they are not expressed at a sufficiently level of abstraction. Many estimations are not computationally feasible if you don’t work at a deep enough level. Documents look prettier in latex and Md. And so on. But I’m willing to concede that there are cases where the above idea holds, if only for the sake of argument.
The point I want to make here is different. My main point is: to think about methodology in purely instrumental ways is a too narrow way of thinking about it even if you only care about it for instrumental reasons.
In the first place, I think it is wrong to think about methods just as instruments for getting results, just as I think it is wrong to think about Rachmaninov’s 3rd piano concerto which is now playing on my computer or the clothes you wear, as, I don’t know, an instrument for socialization. For a long time, and I mean until my mid twenties, I thought classical music was just not for me. Even more, I thought it was just a snobbish posturing attitude. Eventually, I realized that it is a taste you develop and learn. It was at that same time when I started teaching myself baby Rudin and understood that I had been missing a lot of beautiful things.
But even if you are only concerned about how doing methodologically sophisticated work will help you to improve your publication rate, there at least two other reasons. As I said recently , learning to write code is not just about learning data analysis. In my experience, it teaches you many side skills, they incorporate implicitly know how. Since most problems you encounter in research are hard to anticipate, having a broad knowledge in these areas is likely to -forgive the expression- teach you how to think abstractly.
There a is finally, more provocative, argument point I would like to make. Something that is rarely appreciated is that academic communities are, to a large extent, cultures. This division is particularly strong between the humanities and the “sciences”. I use the word culture for a good reason. Many standards are just arbitrary, they are just common practices, social rules, “linguistic” issues. They are just a function of the common problem that each community faces in their everyday practice, but also of just the random, idiosyncratic evolution of those communities over time. This is implicit in many accusations of, for example “celebration of complexity” that people in my current discipline (PoliSci) make against those in my former one (Econ): why express something in a formal model when -and this is often, but not always true- it could be expressed in an easier way in an informal qualitative way?. Sometimes this is justified; sometimes it’s just a matter of common practice, “standards of quality” that everyone is supposed to fulfill.
But, the fact that they do not have an instrumental raison d’être for each single problem, does not mean that these practices lack any sense. The point is that they facilitate communication and to structure interactions. Surely, they are far from being “neutral”, they create a social hierarchy. But very often, they allow communication and coordination which is the engine of the division of labor and knowledge transmission that is essential to research and academia.
Let’s go back to computer science. If you look for “unix developper” in google images, it is easy to see that these were a bunch of very smart chaps, yet maybe not the most concerned about user friendliness in the world. Yet, to be able to communicate with them, you need to understand their language, their practices, and so on. If you get there, you may actually discover that they not fully arbitrary, indeed.