Simulation is a third research modality, alongside mathematical analysis and experiments with real domain entities. The complexity of human subjects makes simulation particularly promising in the social sciences. This complexity often makes mathematical models intractable and burdens experimenters with high costs, ambiguous results, and regulatory hurdles. This book, which exemplifies the simulation approach, originated with a 1993 AAAI workshop on artificial intelligence and theories of groups and organizations. It includes a few of the original workshop papers and several new ones. Taking a computational organization theory perspective, it views computation not only as a method for experimenting with organizations, but as a model for what organizations do.

The editors’ introduction briefly reviews each of the contributions and motivates their assignment to three main  sections. 

The first three papers show the power of the multiagent paradigm for describing organizations. Chapter 1, “Webbots, Trust, and Organizational Science,” by Carley and Prietula, examines the implications of extending traditional organizations with automated information agents. A Soar model of Webbot societies examines performance differences along two dimensions: the size of the society (from one to five) and whether the Webbots are uniformly honest or uniformly dishonest in their collaborations. Interesting variations in performance parameters emerge as Webbots learn and react to the veracity of their colleagues. Chapter 2, “Team-Soar: A Model for Team Decision Making,” by Kang, Waisel, and Wallace, again uses Soar to model an organization, this time four naval commanders in a carrier task group who must access various categories of data to assess the likely threat posed by an unknown aircraft. One study varies the team decision-making method and the amount of information available to each team member in order to study decision deviation and disaster rate. Another manipulates the amount of information, and metaknowledge about which team members have access to which knowledge sources, to explore team efficiency. Chapter 3, “Designing Organizations for Computational Agents,” by So and Durfee, demonstrates organizational self-design in the context of network management systems. The authors build on a model of organizational performance that includes the structure and behavior of the organization itself, the task environment with which it is presented, and the various performance measures of interest to it.

The next three papers explore the relation between organizations and external conditions. Chapter 4, “The Choice between Accuracy and Errors,” by Link, explores the relationship between organizational structure and the kinds of errors an organization can make, using the same problem domain explored in chapter 2. This paper shows how tuning the correlation between an organization and its environment can increase the accuracy of solutions to problems addressed under low time pressure while actually increasing the error rates at higher time pressure. Chapter 5, “Fluctuating Efforts and Sustainable Cooperation,” by Huberman and Glance, studies the effect of nonconstant individual contributions toward a social good and shows how the emergence of bursts of defection can lead to an average utility that decreases in time, even though most individuals are contributing to the cause most of the time. While most studies in the volume use agent-based modeling, this one models its domain using differential equations. Chapter 6, “Task Environment Centered Simulation,” by Decker, illustrates the power of the task analysis, environment modeling, and simulation (TAEMS) framework for modeling organizations. TAEMS can represent relationships among agents of arbitrary complexity and distinguishes three different models of the problem: a generative model that generates problem instances according to specified statistical patterns, an objective model of an actual problem instance, and a subjective model of what individual agents perceive of the problem instance. The author illustrates the framework by applying it to four tasks: going to the post office, hospital scheduling, airport resource management, and Internet information gathering.

The final section contains four papers exploring the impact of information technology on organizations. Chapter 7, “An Organizational Ontology for Enterprise Modeling,” by Fox, Barbuceanu, Gruninger, and Lin, focuses on the competence of a model in answering questions. The authors formalize the ontology underlying the model in first-order logic as Prolog axioms, then test its competence by posing the questions in Prolog and seeing whether the ontology is sufficient to support the proof of the questions. Chapter 8, “Modeling, Simulating, and Enacting Complex Organizational Processes: A Life Cycle Approach,” by Scacchi, discusses the issues that arise in modeling processes interrelated in time, as in the life cycle of a manufactured product. Chapter 9, “An Approach to Modeling Communication and Information Technology in Organizations,” by Kaplan and Carley, shows how simulation can assess the impact of an information technology, such as wearable computing, in the execution of a complex manual task. Chapter 10, “Organizational Mnemonics: Exploring the Role of Information Technology in Collective Remembering and Forgetting,” by Sandoe, explores three models of organizational memory: hierarchical (in which information is stored statically in policies and enforced top-down), network (in which information is stored in the relationships among participants), and hub (in which information is stored in shared information objects).  Sandoe  studies the impact of turbulence in the environment and turnover in the organization’s membership on each of these organizational structures. Hierarchies are the most sensitive to turbulence and the least sensitive to turnover; networks are the most sensitive to turnover and the least sensitive to turbulence; and hubs occupy a mediating position.

The concluding chapter, by Burton, “Validating and Docking: An Overview, Summary, and Challenge,” reviews each of the papers from the perspective of a validation model that includes the purpose of the model, the chosen computational mechanisms, and the experimental design. Burton recalls Axtell’s notion of testing a model by docking it with other models of the same domain that use a different methodology, and suggests that the models in this book might be docked with one another to yield increased understanding.

This volume goes beyond many collections of papers in offering both an integrated bibliography and an index, increasing its usefulness as an integrated reference. Unfortunately, some of the papers referenced in individual chapters are not in the bibliography. Some of the missing references are supplied on a list maintained on the Web by the second editor.