Multi-Agent Systems for Simulation in Geography: Moving Towards an Artificial Geography

International audienceRegularity and persistence expressed beyond contingency are the primary concern of a modeler. The geographer, particularly those specializing in spatial analysis and quantitative geography, might be familiar with this point of view: more than the specificity of individual behaviors, it is the general tendencies beyond the "noise" of this diversity which interests the modeler. These models often sum up and describe observations established by a given scale fairly well. They are nevertheless limited in their capacity to express the conditions for such phenomena appearing. Linking with local dynamics, complexity offers concepts and resources to link global descriptions and spatial analysis. Thinking in terms of complexity implies, among other things, conceiving observed phenomena at the level of numerous interactions which occur between elements operating at one or more lower levels. This position is not recent; it is formulated through systems theory [BER 68], which conceives systems as a collection of objects, of subsystems in interaction. However, system theory, at least in its non-adaptive and applicative form [FOR 80], is more concerned with the modalities of a system function and its behavior than with emergence conditions and their possible evolutions. Chapter written by Eric DAUDE