POEC 7381 MULTIAGENT SIMULATION IN SOCIAL AND POLICY ANALYSIS
Class # 23550
Murray J Leaf U T Dallas Spring 2011
7-9:45 Tuesday FO 2.410
Preliminary syllabus; this will develop.
Last update 13 Dec 2010
Multiagent modeling is a way of using computers to calculate consequences of situations in which multiple actors make interacting decisions according to their own individual rules. It is, therefore, a way of precisely computing outcomes in situations which are non-linear and usually not determinate. Since most causal relations that affect human beings are non-linear and since most situations in which human beings make decisions are not determinate, it is a method of enormous potential importance.
Many proponents of multiagent modeling argue that one of its main values lies in showing how seemingly very complex collective outcomes can arise from very simple rules. Analyses used to make this kind of a point are often either data-free or very data-thin. In some fields, most notably Political Science, almost all the literature is of this sort. Our aim will be work toward building models that incorporate real-world data, and therefore can be used to simulate, and test our analysis of, real-world analyses. This is what some of its proponents call “high fidelity modeling.”
The course will mainly involve learning and adapting NetLogo. This is a widely used ABM software package available free on the web, from Northwestern University. Uri Wilenski is the developer, and is continually improving the language. You can download it from: http://ccl.northwestern.edu/netlogo/ the website also provides a manual that includes a good introductory tutorial, and a user’s guide. These, too, should be downloaded. There are also several help users websites. We may also use a somewhat similar language, called Repast.
My basic plan for the course is that you will first go through the manual and do the tutorials. This will give you a good basic understanding of the programming. It will also leave you with a couple of simple models we can adapt further. This is mainly model of wolf-sheep predation. To develop it further, we will consider how to turn it into a model of managing a common pasture. From here, depending on your interests, we will pick some other models that the Netlogo website provides and see how to adapt them. By “adapt” I mean adjust them so they are more descriptive of real situations, and require real data. For example, Netlogo has a model of self-segregation at social gatherings. It can be adjusted to represent segregation in other situations, and also to represent the kind of segregation that we call "romantic love" that results in male-female pairing.
While we are working here, a colleague who is much more computer-knowledgeable than I am will be doing something similar at the University of Kent, Canterbury, and will be putting some of his students’ work up on a website we will be able to share. I would like to be able to have something good to trade with him. He is planning to use Repast, which is what may lead us to learn some of it.
I will look for a usable textbook, but so far everything I have seen is too generic to be really helpful, and usually not aimed at the problem of making models that are empirically accurate.
Entry portal to simulation modeling in anthropology at the University of Kent, Canterbury: http://era.anthropology.ac.uk/Era_Resources/Pages/Projects/simulate.html
The most comprehensive overview of available ABM software and related language tools I know of is: http://www2.econ.iastate.edu/tesfatsi/acecode.htm
Netlogo webpage on citations or uses of netlogo: http://ccl.northwestern.edu/netlogo/references.shtml
Youtube netlogo introduction: http://www.youtube.com/watch?v=XJ-gO-yAwHU&feature=related
Yahoo Netlogo users group: http://groups.yahoo.com/group/netlogo-users/messages
ASU tutorial for netlogo swidden farming model: http://www.asu.edu/clas/shesc/projects/medland/documents/computermodeling.pdf
The main website for RePast, a more powerful language than NetLogo but also one that is more difficult to learn. We will not use it, but Michael Fischer at Kent probably will: http://repast.sourceforge.net/
Good article in Economist on what ABM models can do that determinate models cannot: http://www.economist.com/node/16636121
Article using ABM model simulating the conflict in Darfur.
Article by Cathy Small describing how she used a simulation model to test different theories about social change on Tonga: http://jasss.soc.surrey.ac.uk/2/3/6.html The article includes a link to download her model, which is written in C++, but which we cannot open up.
Very important article using ABM to simulate prehistoric settlement patterns in the American Southwest: Axtell R. L., Epstein J. M, Dean J. S, Gumerman G. J, Swedlund A. C, Harburger J, Chakravarty S, Hammond R, Parker J, and Parker M. 1999. “Population growth and collapse in a multiagent model of the Kayenta Anasazi in Long House Valley.” Proc Natl Acad Sciences U S A. 99(Supply. 3): c
Another from the same projects as the previous:
The evolution of social behavior in the prehistoric American Southwest (second author; with G.J. Gumerman, A.C. Dean, J.S. Dean, and J.M. Epstein). Artificial Life 9(4): 435-444.
Youtube Netlogo model to simulate crowd dynamics at Burning Man Festival in Santa Fe, by Redfish Group: http://www.youtube.com/watch?v=GcqkeS7jiU8
Website introducing "Digital Archaeology" program at the U of Chicago and Argonne National Laboratory: http://oi.uchicago.edu/OI/PROJ/MASS/papers/DigitalArchaeology.pdf Argonne Labs and the Santa Fe Institute are major centers of development for multi-agent modeling and software systems.
The grade will be based 30% on the class presentations, 30% on the midterm, and 40% on a final paper. The midterm will probably be take-home, essay format, and call for critical evaluation of important theoretical ideas from the readings and discussion.
Course & Instructor Policies
I do not allow “extra credit” or make up work. You are expected to complete all assignments on time. Anything not handed in on time is failed, unless you have made an arrangement with me in advance.
No Field Trips
Standard UTD policies are procedures for all classes are on the U T Dallas website at: http://coursebook.utdallas.edu/syllabus-policies/.