Reviewed by
Marco Remondino
e-business L@B, University of Torino
The main purpose of this book is that of presenting an overview of current work in agent organization. In fact, even if the main title is comprehensive and generic, the topic behind most of the chapters in the book is that of organization, in its widest sense, when applied to multi agent systems (MAS), and this is underlined by the subtitle.The editor is Virginia Dignum, well known researcher in the field of MAS and assistant professor in Information Science - group Cognition and Communication - at the Institute of Information and Computing Sciences at the Utrecht University. As stated in her preface, "organizations in MAS can be understood as complex entities where a multitude of agents interact, within a structured environment aiming at some global purpose". While the goal is immediately clear, the chapters composing the book are not directly linked among them; instead, the book is based on a collection of papers, coming from different authors, constituting the individual chapters, each of them dealing with some aspects of organization applied to MAS. When dealing with this kind of research reports, it gets crucial to group the papers into some main areas; this is done by identifying six sections, each focusing on different aspects of the same topic, not only from the semantic point of view, but also from the technical one. While delving into individual contributions is obviously beyond the purpose of this review, a brief overview of these sections will be given, due to the different contents and backgrounds of the papers.
After an interesting and comprehensive introduction - a paper by V. Dignum herself - which semantically and theoretically discusses the differences among organizing MAS (organization as a process) and MAS for organizations (organization as an entity), the first section, entitled "Methodologies and Frameworks for Agent Organizations" is composed of four papers/chapters about theoretical models. Organizational models are considered as vehicles to express formalizations for agent organizations, i.e.: representing the architecture by using the organizational model itself. For this purpose, some existing meta-models are described in chapters II and III and others are introduced, often based on - and derived from - UML. In particular, in chapter IV, OMACS is epistemologically introduced and operatively described, as a meta-model for designing complex, adaptive systems that can reorganize at runtime, when facing a dynamic environment. In this section MAS are seen as a whole, but underlining their main dimensions, e.g.: agents, environment, interactions, organizations and institutions, using an integral vision. Interactions are seen as a central theme for organization, and in chapter V "Hermes" is described as a methodology to increase flexibility and robustness regarding this aspect.
The second section, entitled "Formal Approaches for Agent Organizations" is more operative than the first one, focusing on the formalisms for modeling and analyzing organizations, illustrated by examples and case studies. Speaking of which, a declarative language for describing agent societies is introduced in chapter VII, while the following paper investigates how organizations can be represented as graphs endowed with formal semantics for links and structural properties. And yet, a model based on Logics is described in Chapter IX.
The third section moves from the formal aspects considering the MAS as a whole, and focuses on agents and their interactions as a central topic. In chapter X, a really interesting and challenging discussion is carried on by C. Castelfranchi, presenting organizations as a macro-micro notion and device that presupposes individual agents at a micro layer, with some level of autonomy and pro-activity. This can be seen as a transitional chapter, from the first part of the book, to following chapters in the section. In fact the author's claim is that MAS discipline should aim at modeling human organizations but, on the other end, it should design artificial organizations for supporting human work. Since the purpose of the paper was, declaratively, not to solve any of the mentioned problems, but to signal them, it puts many challenges and open issues to the community. Other papers in this section aim at modeling organizations through commitments and expectations (XI) and describe agent aiding of human teams to improve performance about communication and self-organization (XII). The final paper of this section considers how agents can adopt organizational rules into their reasoning process. The authors use human organizations as inspiration, by considering an organization as a description of roles, relations and interactions to achieve certain coordination. They propose a modular reasoning model to make organizational rules explicit.
Section four is about the institutional view of the organizations; here, they are seen as regulated by norms and institutions. Once again, in chapter XIV, MAS are seen at a macro level, and described through a logic based meta-model. The next chapter is central for this section, since it's about organizations regulated by institutions, analyzed starting from the institutional concepts themselves, verified by means of "model checking" technique of agent organizations. In chapter XVI, a programming language is introduced, to implement MAS without and with norms, with particular attention to enforcement mechanisms behind institutions.
The fifth section is possibly the most heterogeneous one; it's about organizational dynamics of MAS, and this is a topic that could be faced in many different ways. In chapter XVII a formal model is presented, aimed at studying the organization and dynamics of MAS. Chapter XVIII presents a very interesting policy to dynamically allocate tasks and resources, on the basis of their criticality, between a series of MAS organizations, by adapting to load variations. Experimental results are given, in which the behavior of the system was studied by varying the load of different organizations in MAS. Back to the norms as a central theme, the following chapter treat them in a dynamical way focusing on a transition protocol, as opposed to the static specification of the argumentation protocol. A simulation is presented in the final chapter, whose aim is to understand the effects of formal and informal structures on the performance of organizations.
The closing section is about applications, presenting agents as personal assistants (XXI), a method for designing organizations at run time (XXII) and a Petri Net based model for monitoring and controlling each level of an organization (XXIII).
The heterogeneity of the chapters, grouped under the macro theme of organization in MAS, is the main point of strength of the book. The fact that the contributions are not directly linked among them makes the book worth reading for a wide range of scientists and practitioners in the field of agent based systems. At the same time, since the book includes so many different papers, dealing with so many subjects in the field of MAS could result sometimes confusing, since even within the same section some works are heterogeneous for which concerns the scientific approach, methodology, and even terminology, thus somewhat resembling a conference book of proceedings. Besides, since this is a handbook of research about Organizational Models in MAS, many chapter are indeed ongoing researches, so many open issues are exposed. Definitely not a guide for beginners, the book is intended for researchers in some field of MAS, desiring to explore and discover others.
Return to Contents
of this issue
© Copyright Journal of Artificial Societies and Social Simulation, 2009