To understand agent based modeling (ABM), there is an idea of describing a system from the various and distinct perspectives of the units that make up the system, rather than vewing the model as a technological tool. The agent based model can be seen as a set of differential equations that separately describe the dynamics of each particular system component. An alternative would be to just dive in and to start describing the system without respect to the perspectives of each component.
Each description of a system component is referred to as an “agent”.
ABM has benefits that include flexibility along with the ability to identify emergent phenomena, and to give a natural description of a system. Emergent phenomena is viewed from the perspective that there is more to complex interactions than the description of activities in a part of the system. The interactions result in properties that are unrelated to the properties of the parts of the system, leading to “a whole is more than the sum of parts” way of thinking.
A confrontation between two groups of demonstrators in a mass of demonstrating groups, may take demonstrators from both groups in opposite directions from their original route, such as to the hospital or jail. The complex interaction that caused the confrontation resulted in an emergent phenomena of fighting, injuries, interaction with law enforcement and emergency response personnel and ultimately, completely away from the scene.
The natural nature of ABM involves letting people tell what it is that they actually do, rather than using equations that govern or describe their behavior.
An example would be to lay out a virtual museum exhibit, give visitors a guide to the various parts of the exhibit, then let them choose where to go. From the perspective of a component of the exhibition (people), facts about the most popular parts of the exhibit and the order of peoples movements would constitute natural modeling as opposed to using pre-existing ideas of people’s behavior in exhibits The resulting information would give the curators much more natural information about the popularity and attraction of the various components of the exhibit before they go through the expense and activity of putting it all together in the real world.
Finally, ABM is flexible enough to allow for changing, adding and moving agents. Each agent can be reconfigured with differing complexity, behavior, degree of rationality, evolutionary ability, learning ability and rules of how the agent interacts. Agents can be aggregated into groups that can become subgroups of larger groups, and so on. Agents can have differing descriptions, with any level of complexity and can have multiple levels of complexity within the same model.
In other words, ABM allows for adustment, tweaking and adapting as new things are discovered or as more facts allow more complexity in description, interaction and behavior to be introduced.
The major applications for ABM right now include understanding: traffic flows, market behavior, organizational behavior and something that is called the diffusion of innovation and adoption dynamics. The final category involves systems where people are influenced by whatever it is that others around them do. This is a new, cutting edge and as yet not widely usable component of ABM.
Eric Bonabeau, “Agent Based Modeling: Methods and Techniques For Simulating Human Systems”, National Academy of Sciences, 2010