O’Sullivan gives a very good survey about the agent-based models (ABM) in spatial science research. He begins with different definitions of ABM, analyzing their advantages and disadvantages. Then he categorizes ABM applications into three types, with respect to their degree of complexity. As mentioned by Bonabeau, ABM provides an efficient approach to describe the complex systems. O’Sullivan illustrates this point in the paper by delineating spatiotemporal and social ABM representation methodologies. In geography research, ABM provides a powerful tool for modeling geospatial information on computers. However, the challenges should not be overlooked. As ABM represents geospatial information at individual level, the complexity and model verification are becoming more difficult with the increase of agent numbers.
By modeling interesting entities as agencies, ABM reviews the relationship and interactions between these agencies. Sometimes, the study targets in GIS research are in a great number and their relationship can be very complicated for statistical modeling. But ABM can achieve that by modeling each entity or each attribute of the study target, providing detailed models about the intricate phenomenon.
I propose that ABM should be integrated with other geospatial analytical methods in GIS research. By viewing a large number of agencies at aggregate level, we can find several interesting discoveries that cannot be reflected by studying the interactions between the agencies. For example, geospatial statistics can study climate changes at global level, by utilizing geospatial information provided by a large body of agencies which contribute to the climate change. Therefore, utilizing other analyzing tools with ABM can help us in GIS research.