Information Technology Reference
In-Depth Information
Nevertheless, the following benefits compared to other modeling
paradigms are frequently associated with agent-based modeling and
simulation:
Agent-based modeling allows a natural and often very intuitive,
structure preserving modeling and implementation of the system
under investigation [68, 22, 83, 27]. Furthermore, it is often more
easy to describe the behavior of single agents and provide necessary
data than to describe the behavior of the overall system.
Agent-based modeling has the power to demonstrate emergent
phenomena which are fundamentally a multiresolution concept. In
other words, agent-based modeling allows a comparatively easy
integration of different levels of modeling and observation [68, 15,
120], [84, 113]. This fits well to the fact that agent-based model
development (and calibration) is often based on more fine-grained
data as input compared to other modeling approaches [83].
Agent-based modeling is flexible and provides high scalability. The
level of detail within a model can be adjusted quite easily as agent-
based modeling provides a natural framework for tuning the com-
plexity of the agents (e.g., behavior, degree of rationality, ability
to learn and evolve, and rules of interactions). Furthermore, agent-
based models inherently support parallel and distributed simulation
execution and promise a high degree of maintainability as well as
extensibility [22, 83, 27, 23, 68].
In summary, building upon 'proven, highly successful techniques such
as discrete event simulation and object-oriented programming, agent-
based modeling leverages these and other techniques as much as
possible to produce returns more directly and quickly' [93, p. 5f.].
Although the idea of agent-based modeling and simulation is easy
to grasp and understand, and an implementation is technically quite
simple, two major drawbacks can be identified [68, 60]:
1. The relation of behavior on the micro-level (i.e., of single agents)
to behavior of the overall model is extremely dicult to predict
