ABMS and GIS
Complexity Science Research
Solving Problems Using Techniques in Spatial Science
Tutorial for using Repast with GIS
There are simple and there are complex systems. Simple systems have cause and effect linkages that are obvious. Not every system can be explained by a formula or an x-y graph. Some systems at the surface appear chaotic, but beneath the surface show patterns and properties that emerge incognito. Some systems need to be approached by studying how the relationships between the parts give rise to the overall patterns of the system and how the system interacts and forms relationships with its environment. Ecological systems often have so many actors and factors interacting that they cannot be described by a formula, yet even these systems demonstrate collective patterns. These systems are called complex systems. Scientists and researchers have gone to great lengths to explain even the most simple systems. The world's complex systems, like simple systems, need explaining.
Emergence explains complex systems conceptually. Emergence suggests that large complex systems such as that of a hurricane, or a body of water, can be explained through combining a multiplicity of individuals interacting with very simple rules for behavior. The patterns of high winds and heavy downpours of rain are seen as a result of a multiplicity of simple atmospheric interactions. The cohesive properties of a drop of water or the ebb and flow of the tides in an ocean can be seen best as the result of millions (or almost infinitely more) of interactions between tiny H2O molecules each with simple properties of attraction and repulsion, and properties such as mass, with which an outside actor may interact. Emergence can also be seen in societies where, for instance, fashion or language trends become popular when they are adopted by certain individuals who have the ability to influence many others (Malcolm Gladwell's book, The Tipping Point talks of these "Connectors").
But how does one model emergence? Until recently, people have not even considered the possibility of modeling these complex systems using technology. Though the computing capabilities have been there, they were too expensive and inaccessible to be of any help to most researchers. Besides not having the memory and the processing power, there were no programs that would allow such simulations to be modeled. Most modeling tools assumed system inputs and outputs resembling mathematical equations, and could not be expected to demonstrate emergent behavior, which is based on individual behaviors.
There is, however, a modeling technique that has been developed initially for the field of computer science that is now being used (click here to see an article and a bibliography of works done on this) to understand complex systems in ecosystems and social systems. The technique allows for the creation of individuals who have memory, can communicate with others, have properties assigned to them that can change throughout the simulation, have rational decision-making capabilities, are affected by their environment, can change their environment, and can adapt their behavior to changes in their environment through learning. This technique is called agent-based modeling. Using agent-based modeling in combination with spatial technologies such as geographic information systems has become a potent and essential technique for many companies, organizations, and government agencies trying to come up with solutions to their complex problems.
The goal of the research is not to create simulations that tell you everything you need to know about a system. No matter how complex the model, there is always some aspect of a system that remains unexplained. The goal is to demonstrate the most significant patterns in the system by modeling only what is necessary to observe those patterns that are present in reality. Often, this is easier than we expect. While an agent's environment determines what it does and where it does it, the agent also acts on its environment. For instance, a young professional purchasing a home in a globalizing city will increase the chances that the costs of the buildings around that home will go up.
I do work with Raja Sengupta and Ben Forest, professors and researchers at McGill University in Montréal on integrating agent-based modeling systems and geographic information systems with approaches to modeling urban residential dynamics. I've built a model in RePast to simulate residential dynamics in a certain region in Boston. The goal of the project is to show whether agent-based modeling can be used to simulate gentrification dynamics, particularly the movement of the group Richard Florida has termed the "creative class" in globalized cities. The thesis explains, assesses and tests the validity of the model by comparing simulated data with real data. If you're interested in my methodology, I've created a tutorial that shows how to integrate GIS with agent-based modeling using RePast.
Here are two ascii files of maps created from the simulation that can be opened in ArcMap or ArcView:
And this is the real 2007 data in ascii format.
Jeremy Jackson Home