Agent-Based Simulation Modeling in Social and Organizational Domains

摘要

Much of the world is composed of phenomena that are usefully represented as discrete and identifiable objects. Further than this, some of the world consists of objects that are usefully thought of as having some sort of cognition in the form of learning or decision-making processes. These include animals (including, of course, humans) but also groups of animals, companies, political parties, and even countries. While some of the phenomena we observe are simple enough that we get acceptable results by aggregating these entities within statistical or numerical models, it is clear that this is not possible with all the phenomena we observe. In other words, it is sometimes necessary to explicitly represent these entities as individually tracked entities within the simulation—resulting in individual-based simulations. When these entities in the model are usefully characterized as having some sort of cognitive processes, we call these entities agents—resulting in agent-based simulations. Thus, in our view, what makes part of a computer simulation an "agent'' is not dependent on how it is implemented or its internal properties but how it is usefully characterized—if it is useful to adopt what Dennett [1] calls an "intentional stance" toward parts of a program, then it is an agent for this is all that this means. An analogy can be drawn with quantum physics—whether one models a photon as a particle or a wave depends on the situation at hand and the goals of the modeler. This view contrasts strongly with other views that attempt to formulate a set of necessary properties of agents [2].