Towards Computational Strategy (Part I)

(by Adam Elkus)

As Zenpundit readers may know from my previous entry, I am a PhD student in Computational Social Science at George Mason University. Though I am learning the technical craft of computer modeling of social processes, I have had a longstanding interest in future war and technology. I grew up in California, and in an environment very much shaped by the technology industry. This was diluted by the fact that I grew up in Southern California and also have had a mostly liberal arts education heavy on arts, (military and strategic) history, and social science. My own struggle these days is — having spent so long developing the humanities side of myself — to think more like an engineer in developing computational and quantitative approaches to studying social science.  You can see some of my notes on this process at my own personal research journal.

My term project for my CSS 600 class is a very, very crude and simple agent-based model of military mobilization. I’m also working on an equally crude model of strategic learning and a very crude simulation of strategic effectiveness in alliances. I don’t like them, and am on the fence about whether I’ll want to post them on Github after all — though I am definitely going to post the alliance effectiveness model (which relies on an interesting optimization algorithm) to a larger audience.

That is OK. Though I began practicing Python and NetLogo all the way back in early spring, learning to program from scratch takes time and effort. Through my classes, tutoring, and plenty of sleepless nights I bootstrapped my way to being able to make computational models in Python, NetLogo, and Java. And this winter I will be practicing Java and Lisp in preparation for spring classes. I could probably, with more time and less distractions (readers who know me in person will know that I unfortunately have had to devote a lot more attention this semester to resolving some logistical problems outside of academics) have done better than the models I’m making for my classes. So I chose easier and simpler for my first models. However, I have grander ambitions in the long term.

This post is the first of a series that I am constructing from notes I have scribbled throughout this semester, my first at GMU. I have, over the last year, relentlessly explored and narrowed down my idea of my research agenda. Aaron Frank, Jay Ulfelder, Mark Safranski, A.E. Stahl, David Masad, Russell Thomas, Lynn Rees, Dan tdaxp, Daniel Trombly, Joshua Foust, Trey Causey, Alex Hanna, Sina K., Anton Strezhnez, Nick Prime, Daniel Bilar, Sam Liles, W.K. Winecoff, H. Lucien Gauthier III, Dave Lyle, Daniel Solomon, Jon Jeckell, Alex Olesker, Brett Fujioka, Robert Caruso, the mysterious Dr. Kypt3ia, and many others too numerous to mention have served as sounding boards for a successive array of both promising ideas and also half and even quarter-baked “dry holes.”

I have a gigantic array of TextEdit files, Moleskine journals, and even theories scribbled in pseudocode in my Sublime Text 2 text editor. So as I turn them into coherent posts, I will space them out individually. This series concerns the concept of “computational strategy,” which I am shaping my own studies around. For example, I will be taking a survey artificial intelligence class next semester — one of two survey courses that computer science majors (which I am not  — I will have the same relationship with CS that political science has with probability and statistics in that I’ll try to borrow as much as I can but also will never be as good as an actual CS student) must take to survey breadth and depth of AI. I will also be taking a course on cognitive programming for computer models.

It will consist of the following posts:

(I) My own journey as a PhD student up to this point

(II) Contrasting generative social science with theoretical computer science — and their deficits when applied to strategy

(III) From “killer robots” to “robot historian” and computation as a universal language

(IV) Towards a preliminary research agenda for computational approaches for studying strategic theory

We begin with (I):

Between A Dead Prussian And Kenneth Waltz

Page 1 of 3 | Next page