Position Paper for Agent-Based Models of Complex Systems

Object-oriented process models, which include individual-based models (IBM) commonly used in ecology and agent-based models (ABM) common in the social sciences, allow for modeling both change and movement of geographic entities and have developed independently of GIS. The object-oriented framework of ABM involves identification of agents and of a temporal framework within which those agents perform actions. While many different types of agents can exist, the following general definition is common: an agent is a self-directed object, i.e., it has the ability to satisfy internal goals or objectives through actions and decisions based on a set of internal rules or strategies. These agents may be dynamic in either state (i.e., change) or space (i.e., movement) and may, through their actions, change the state or location of other objects, processes, or environments around them. Agent dynamics are most naturally implemented in an ABM by a set of behaviors (" methods ") that can include conditional decision making and other (non-linear) rules that distinguish them from mathematically continuous models. The ability for Lagrangian motion (i.e., agent movement) distinguishes ABM and other object-oriented modeling frameworks from the change-based spatial models described above. It also creates additional challenges for integrating these models with GIS, as described in more detail below. ABM dynamics are defined at the level of (a) agent behaviors that result in change and movement, and (b) the independent dynamics, if any, of non-agent objects. Thus to represent dynamics, ABMs are implemented as discrete event simulations, in which some kind of "scheduling" mechanism handles the sequencing of agent behaviors and events. An ABM may implement scheduled events in three ways: • Events may be sequenced in a synchronous step-wise fashion. For example, each agent, set of agents or non-agent object is signaled to perform its tasks once at each time step or once every n time steps. • An event may be scheduled to occur only once at some time step n. Any number of different events may be scheduled to occur in this fashion providing a predetermined history of events to take place. • The model may encapsulate 'event-driven' processes whereby model agents may trigger events to occur or may add events to the schedule or queue of events to take place. On the other hand, ABMs often use relatively limited representations of space. For example, ABMs frequently use hypothetical spaces based on square or hexagonal tessellations, and only recently have …