Complexity in World Politics: Concepts and Methods of a New Paradigm

Complexity in World Politics: Concepts and Methods of a New Paradigm
Edited by Neil E. Harrison

The feeling I got from this book was similar to much of the work I found on complexity and the social sciences, especially those that emphasized a place for agent based models. Complexity, emergence, and agent-based modeling offers an opportunity for greater scientific legitimacy, in this case, for the study of world politics. There is already a long history of modeling in political science, but I think the complexity banner has managed to illustrate that the models aren't just numeric abstracts but have power to explain the strange dynamic world of global politics. I'm going to focus on two chapters in particular.

I was disappointed with this book. While there was a good amount of discussion of the implications of complexity on the study of world politics and the usefulness of agent based modeling, there wasn't any actual modeling! That might be a big plus for the computational modeler - lots of interesting open questions have been posed, get to work! Regardless of the lack of actual models there is still a lot of value in this volume.

One of my favorite chapters was by Matthew J Hoffman. I have read papers by him before and I wasn't surprised that I enjoyed his take on environmental politics. In environmental politics there has long been something called 'regime theory'. Regime theory is supposed to explain how states can reach compromises without some hegemonic central authority to guarantee or enforce the agreements. He focuses on the Montreal Protocol and abstracts 'rules' to the 'agents' of the international system. There are rules like 'CFC control is a North-only problem.' Changes to the environment due to evaluations of the Montreal Protocol by the United States causes the rule to change to 'universal participation' precipitating a change. This is a nice theory, methinks, but it needs a model! Hoffman ends the chapter saying that this explanation will gain great credibility if a model can be produced using theoretical assumptions that produces universal participation as an emergent property. Hm... anyone game? The theory needs to be cleaned up a bit so that a model can be reliably detailed, but it's possible.

Another interesting chapter was by Ravi Bhavnani on ethic norms and violence. I am reminded of one of the open questions Miller and Page posed of complexity science, When does heterogeneity matter? Well, Bhavnani has identified one case: ethnic violence. There seems to be, on the one hand, a normalization of deviance argument. Killing the other group became the norm and a game. As Hoffman, Bhavnani poses a model that isn't actually created. In this case he suggests that the model has no business being empirically validated as such validation is ridiculously impossible. He suggests that agents have varying levels of out-group extremism and in-group favoritism and that agents need a mechanism of punishing slackers - these assumptions are informed by careful analysis of the Rwandan genocide. There, of course, should be mechanisms for learning such as adopting the behavior of those the agent encounters. In addition the agents should have a pre-existing network or an endogenously-grown network of relationships between each other. A number of things could be explored, such as what the structure of the network does, the effect of varied compositions of pacifists and extremists, and what sort of effect the absence of punishment has. This model would give social scientists a laboratory to experiment with the emergence of a cascade of ethnic violence - since experimenting with the real thing is not very feasible. The author ends the chapter by championing agent-based models in contrast to game-theoretic or equation-based models - and I generally agree.

This book poses a lot of great questions, but provides very few answers. I feel that anyone who is curious about agent-based models or complexity in world politics would see a lot of value here. Although, this book is probably best for the agent-based modeler who is out of ideas and wants to publish some low-hanging fruit.