Exploring Complex Systems in the Social Sciences

Greg Marfleet, Carleton College

Since 2004, I have regularly taught a course titled the Complexity of Politics. I introduce students--who are mostly political science and economics majors--to the application of agent-based, computational modeling techniques in the social sciences. The course explores some of the important concepts of complexity as they relate to social and political phenomenon including emergence, adaptive agents, co-evolution, positive and negative feedback systems, non-linear processes self-organized criticality and tipping points, and perpetual novelty. We look, for example, at the co-evolution of strategies among political parties in electoral competition, the emergence of alliances and power-balances in international relations and how feedback can push civil unrest past a tipping point into civil war. We explore these models and others through readings and a workshop-driven, hands-on, model-building experience using NetLogo programming software.

The primary challenges I have encountered teaching this class arise from the fact that Complexity-oriented modes of inquiry are very new to the social sciences. This problem manifests itself in several ways including lack of teaching resources, limited prerequisite skills among students and even accessibility of intellectual and conceptual foundations among students.

First, what is available in the way of teaching resources has generally been oriented toward economists and geared toward graduate-level instruction. Few undergraduate programs offer any training in this area almost none in political science. Consequently finding texts books or even topically-relevant research papers accessible to students has been a challenge. Fortunately this has become easier over the last five years as more research has been presented. However, there is still no complexity text book for undergraduate political scientists. The closest item I have found to a political science oreiented introduction is Axelrods "Complexity of Cooperation" (1997). Over the years I have used Schelling's "Micromotives and Macrobehavior" (1978) and currently use Miller and Page's "Complex Adaptive Systems: An Introduction to Computational Models of Social Life" (2007).

A second hurdle I have had to overcome is a common lack of student familiarity with computer programming. It may seem incongruous that our highly internet-savvy and 'wired' cohorts of students that we have in our classes may be less exposed to the basics of programming than we were in our high-school days. Most of our students have a sufficient background in mathematics and statistics to feel immediately comfortable in their mainstream methods training; remarkably few have any background in computer programming in any language. Since I was determined to incorporate a direct experience in model building into the complexity course, I have had to include many of the elements of an introductory computer science course into the first few workshop assignments. Topics that I address in the first three weeks including algorithmic thinking and the basics of variables and flow control using loops and conditional statements.

The third and most central challenge to teaching this course relates to the way social scientists, particularly political scientists, approach social explanation. For most social scientists, methodological training focuses on statistical techniques for causal inference. Our undergraduates begin their methods education with Gaussian, linear, statistical approaches to hypothesis testing. A key feature of these techniques is data aggregation. When we look at large-sample survey data, for example, it is virtually always through descriptive summary statistics or bivariate and multivariate correlation or regression models. One consequence of this training is our tendency to be more attentive social outcomes than to social processes and to focus on the median indicators.

Political science, in particular among social sciences, has tended to engage in 'top down' (macro) level explanation more than 'bottom up' (micro) theorizing. Institutions, structures and elites draw our attention. Typical explanations for election outcomes, for example, arise from factors like elite campaign behavior, voter registration laws, mode of districting or balloting systems. I have found that economics students, who have been exposed micro-economics and/or game-theory, have an intuitively stronger grasp of bottom-up processes through their ubiquitous market metaphor. A key goal of the early workshops for the class is to encourage political science students to begin thinking about exploring social processes and to conceptualize macro-level outcomes as the result of a series micro-level interactions situated in time and space.

One of the early workshop assignments that I use to do this is the "Standing Ovation Problem". This problem was suggested by Scott Page in a lecture that was provided in a class in complexity and modeling that he taught at the University of Michigan's ICPSR workshops. Based on this lecture I developed a NetLogo assignment that walks students through some basic code and suggests possible expansions. Subsequently, Miller and Page have written a short paper exploring the SOP problem as a teaching tool which I have students read after they have tried to write the program. I have uploaded both the course syllabus and the SOP assignment.

Citations:

Axelrod, Robert. 1997. The Complexity of Cooperation. Princeton University Press: Princeton NJ.

Miller , John H. and Scott E. Page . 2007. Complex Adaptive Systems: An Introduction to Computational Models of Social Life. Princeton University Press: Princeton, NJ.

Schelling, Thomas C. 1978, Micromotives and Macrobehavior. Norton, New York.

Wilensky, U. (1999). NetLogo (more info) . Center for Connected Learning and Computer-Based Modeling. Northwestern University, Evanston, IL.