Description

STAR is a system for modeling and simulating multi-agent organizations under a wide variety of tasks. The system is composed on four core components: Structure, Task, Agent, and Resource. Through these components, and features of STAR, modelers and researchers can rapidly prototype and test organziations of agents performing tasks. These features include a GUI-based modeling system, two simulation engines, advanced organizational statistics, an Application Program Interface (API) to add new agents, Import and Export of standard organizational matrix data, and a turnover tool.

Motivation

Organizations are performing major reorganizations of their divisions and employees. These reorganizations are often designed by experienced professionals: consultants outside the company or executives currently employed by the company. The reorganizations are complex and promise either increased performance or a return to profitability. However, these reorganizations are usually performed in an ad-hoc manner. While experts are utilized, those experts do not have access to tools that allow them to systematically explore alternative reorganization plans.

Contributions

STAR provides four major contributions. The first is the system itself as an expressive multi-agent modeling and simulation system. This expressiveness is far more than allowed in extant systems, and has been shown through the modeling and simulation of three representative tasks: Garbage Can Model, Radar Task, and a Job Shop Task.

Second, STAR introduces the concept of Permeability. Traditionally, organizational modelers have expressed the boundaries of organizational models as fixed or non-permeable. That is, a model of agent proposed by some researcher is closed; no part of an agent could be modeled elsewhere, only in the proposed model. Examining multiple models of agent, task, and resource, it become clear that there is no singular way to define a model. Indeed, it is apparent that boundaries are permeable. For example, some researchers regard knowledge as a resource. Others regard it as a feature of an agent. Knowledge, then, is permeable. STAR supports permeability by allowing modelers to exploit the flexibility of the STAR system.

Third, STAR introduces a new method of Morphing organizations. Applied to organizations, morphing is a method to study organizational change. The method allows the relationships between model components (i.e., structures) to change through time according to a set of simple rules. At a more general level, morphing may be an approach to permeability. In addition to relationships between components changing, the parts of components may morph from one component to another.

Last, STAR introduces a concept of Composite Agency. A composite agent is an grouping of existing agents into a single unit. The single unit is then treated as an agent unto itself. Composite agents can be used to first model the micro behavior of organizational subunits. Then, creating composite agents of each subunit, the entire organization can be simulated. Composite agency can also be used to more easily test new theory on top of existing agents.

Working Paper

Kaplan, D.J. (1999) "The STAR System: A Unified Multi-Agent Simulation Model of Structure, Task, Agent, and Resource," CASOS Working Paper, Carnegie Mellon University.