Developments in Intelligent Agent Technologies and Multi-Agent Systems: Concepts and Applications

Developments in Intelligent Agent Technologies and Multi-Agent Systems: Concepts and Applications

Goran Trajkovski (Algoco eLearning, USA)
Release Date: November, 2010|Copyright: © 2011 |Pages: 396|DOI: 10.4018/978-1-60960-171-3
ISBN13: 9781609601713|ISBN10: 1609601718|EISBN13: 9781609601737
List Price: $180.00
20% Discount:-$36.00
List Price: $180.00
20% Discount:-$36.00
Hardcover +
List Price: $215.00
20% Discount:-$43.00
(Individual Chapters)
No Current Special Offers


Agent technologies and multi-agent systems is an emerging field of study that is unique in the sense that all efforts to build this discipline unite knowledge from different areas of study.

Developments in Intelligent Agent Technologies and Multi-Agent Systems: Concepts and Applications discusses research on emerging technologies and systems based on agent and multi-agent paradigms across various fields of science, engineering and technology. This book is a collection of work that covers conceptual frameworks, case studies, and analysis while serving as a medium of communication among researchers from academia, industry and government.

Topics Covered

The many academic areas covered in this publication include, but are not limited to:

  • Agent models and architectures
  • Agent-Based Modeling and Simulation
  • Agent-based social simulations
  • Agent-oriented software engineering
  • Agents in Electronic Business and Virtual Organizations
  • Artificial social systems
  • Colonies and swarm intelligence
  • Conceptual agent frameworks
  • Developmental and cognitive agents
  • Multiagent Systems

Reviews and Testimonials

This book is a compilation of the articles published in the four issues of the first volume of the International Journal of Agent Technologies and Systems. It showcases multiple (personal) researchers' journeys coming from many directions and traditional disciplines and backgrounds, to help advance the emerging science of Agent Technologies and Systems.

– Goran Trajkovski, Algoco eLearning, USA

Table of Contents and List of Contributors

Search this Book:


This book is a compilation of the articles published in the four issues of the first volume of the International Journal of Agent Technologies and Systems (IJATS). (Trajkovski, IJATS, 2009). It showcases multiple (personal) researchers’ journeys coming from many directions and traditional disciplines and backgrounds, to help advance the emerging science of Agent Technologies and Systems. 

Anything surrounding us could serve as a motivation for a study or exercise in agents or multi-agent systems. That is what I, as well as many of my colleagues find fascinating about what we do. For example, if you have an infant or own a pet, you are probably fascinated when watching them learn new environments and interact with them and other agents in these environments. The motivations in my personal research at its early days, for example, came from being intrigued in watching infant explore of an environment and watching humans interact with their pets. I became fascinated by the emergent behaviors in agents and their societies, regardless whether they are artificial or biological, homogenous or heterogeneous. How people think, make decisions, communicate, and imitate each other are questions that have always been challenging to me. Everyone else that researches these areas, I believe, has their personal story from where the passion for studying this emergent cross/interdisciplinary area stems. 

This was the motivation to establish the International Journal of Agent Technologies and Systems in 2009.

In this section we will review the founding principles behind International Journal of Agent Technologies and Systems that have been shared with the scientific community in the wide campaign of soliciting contributions. 

International Journal of Agent Technologies and Systems publishes original contributions in the areas of theories of agency and multiagent systems. It covers conceptual frameworks, case studies, empirical analysis, analytical and simulation models of agent anthropologies and sociologies, and their application. It covers topics that include the following general areas:

• Conceptual agent frameworks 
o Agent development environments   
o Agent models and architectures  
o Representation of agents and representation in agents  
o Modeling other agents and self  
o Developmental and cognitive agents  
o Knowledge management and ontologies 
• Simulations and constructions of agents 
o Agent-based modeling and simulation  
o Tools and cases  
o Agent-based social simulations  
o Emergent behavior in agents and agent societies  
o Multi robot systems  
o Robot teams 
• Multiagent systems 
o Learning in multi-agent systems  
o Social and organizational structure in agent societies  
o Inter-agent interaction  
o Agent languages   
o Information propagation and exchange in multi-agent systems  
o Artificial social systems  
o Homogenous and heterogeneous agent societies  
o Colonies and swarm intelligence 
• Applications 
o Human-agent interaction  
o Interface agents  
o Virtual humans   
o Software and pervasive agents  
o Agent-based data mining  
o Agent-oriented software engineering  
o Agents in electronic business and virtual organizations  
o Ethical and legal issues pertaining to agency and multi-agent systems.

Naturally the topics are not intended to act as requirements but rather as examples of acceptable topics that are of interest to IJATS. This volume illustrates a diverse range of questions that were taken upon in the first volume of the journal.

The area of Agent Technologies and Systems is an emerging field of study, and, arguably, does not yet have a mainstream core that the majority of researchers agree on, or build their research on. All efforts are unique in the sense that they explore and unite knowledge from different other disciplines in contributing to what has been an exciting emergent new field and discipline.  Journals, such as IJATS, and volumes such as this one, are themselves acting as agents itself in building the bona fide science of Agent Technologies and Systems.

IJATS, in my subjective opinion, hit its goals from its very first issue. It was set to increase awareness and interest in agent research, encourage collaboration and give a representative overview of the current state of research in this area. It aims at bringing together not only scientists from different areas of computer science, but also researchers from different fields studying similar concepts. The journal now serves as an inclusive forum for discussion on ongoing or completed work in both theoretical and practical issues of intelligent agent technologies and multi-agent systems. It focuses on all aspects of agents and multi-agent systems, with a particular emphasis on how to modify established learning techniques and/or create new learning paradigms to address the many challenges presented by complex real-world problems. IJATS was created this journal to disseminate and discuss high quality research results on emerging technologies and successful systems based on the agent and multiagent paradigms, with a comprehensive coverage and understanding of the implications of these paradigms from and to various fields of science, engineering, and technology. It is an interdisciplinary journal that brings together researchers from academia, industry, and the government in discussing conceptual and implementation issues in using the agent approach in solving real life problems. It also acts as a medium of communication among those researchers and practitioners with interest in exploring the benefits of the concepts, simulations, constructions, and applications of theories of agency beyond disciplinary boundaries. It is an open forum for exchange of ideas, so that neither of us reinvents the wheel. Often times the solutions we are looking for has be found in a different discipline in a different context.   

Deciding the order of the articles in an issue of the journal can sometime be tricky in a discipline that is just emerging and stabilizing. Putting 19 articles together in a book is a significantly more complicated task. The order of the chapters needs to tell a consistent story with a logical flow. 

Our Handbook of Agent-Based Societies: Social and Cultural Interaction (Trajkovski & Collins, Handbook of Research on Agent-Based Societies: Social and Cultural Interactions, 2009) is the natural prequel of this book. This Handbook was organized in three sections: Initial States, Emergencies, and Second-Order Emergencies. The first part, Initial States contains the chapters that refer to the foundation concepts that are treated with models in the other sections. The second part, Emergencies, contains models of artificial societies where the agents interact in the environment and phenomena emerge, but unlike in the chapters in part III, Second Order Emergencies, the agents do not change their behaviors based on these emergencies. (Trajkovski & Collins, Towards More Lively Machines, 2009) This creates a platform for classification of research in this area.

In this book, we continue the same tradition from the Handbook, while taking into consideration the papers that were present in IJATS in its first year. The following three (loosely distinct) sections of the book emerged: Concepts and Challenges, Emergencies, and Applications. Naturally, there are no clear-cut topical boundaries, and chapters could easily migrate between sections, or be reordered.

As Bragin states in his review (Bragin, 2009) of the Handbook of Agent-Based Societies: Social and Cultural Interaction (Trajkovski & Collins, Handbook of Research on Agent-Based Societies: Social and Cultural Interactions, 2009):

“Agent-Based Modeling is generally held to be the core methodology of human complex systems science. Although it is possible to express any ABM in mathematical terms (see (Erdi, 2008), § 9.1.3 for an example of how this has been done with one of the Sugarscape models of Epstein and Axtell) there are a number of reasons why ABM approaches are fundamentally more useful, practical and provide greater explanatory insight for social scientists than do mathematical and statistical models (see (Macy & Willer, 2002) and (Gilbert, 2008)). Nevertheless, the agent-based modeler must be as rigorous as the mathematician or statistician: "agent-based models have to be complete, consistent, and unambiguous if they are to be capable of being executed on a computer" (Gilbert 2008).”

It is exactly with these challenges that we open the book with. The first chapter investigates the notion and nature of time in ABS, and stresses how important time considerations are in being able to interpret results from simulations. Koehler’s chapter “Attending to Temporal Assumptions May Enrich Autonomous Agent Computer Simulations” (Koehler, 2009) is a study that transcends computer science and philosophical ontology. It examines how agents and environments are oriented in time and this orientation’s relevance to observing and interpreting emergent phenomena. This study emphasizes the problems in interpreting computer simulated results of agent societies, and the difficulties in deciphering what exactly those digital simulations show when investigating an essentially analog phenomena in a world evolving in time. The essay begins with the observation of the NetLogo platform, a programmable simulation computing environment. (Wilensky, 2009) and the temporal assumptions in common computer stimulations. These discussions have long been a topic in the studies of non-linear systems, sensitive to initial conditions, where generally the concept of chaos was studied at length. Koehler digs deep into the J.T. Fraser’s the hierarchical theory of time (Frasier, 1998) that distinguishes between five types of temporality. He argues that the proposed temporal deepening of how simulations are constructed involving interaction of multiple temporalities could lead to the unexpected triggering of an avalanche of unexpected phenomena. The article raises important questions on the validity of the conclusions we get out of stimulation explorations, juxtaposed to philosophical (and other) concepts and understandings of time.

The following chapter expands the Game of Life by experimentation of the timing in the simulations, and more. In their paper “Agent Interaction via Message-Based Belief Communication,” Conover and Hammell (Conover & Hammell, Agent Interaction via Message-Based Belief Communication, 2009) reflect into the concept of “temporally autonomous” multi-agent interaction. This paper builds upon the problems with discrete simulations. Namely, in most simulations, agent behavior is regulated by a global timer where all agents act and interact deterministically in time. The authors argue that this discrete timing mechanism yields an artificial reflection of actual physical agent interaction. To this end, a specialized simulation framework is developed, several simulations are then conducted from which data are gathered and it is subsequently demonstrated that manipulation of the timing variable amongst interacting agents affects the emergent behaviors of agent populations. The observations of agents interacting “passively,” (Conover & Trajkovski, Effects of temporary asynchronous interaction on simple multiagent behavior, 2007) are now observed in simple “active interactions.” The authors observe that the timing of the agent interaction influence the emergent behaviors in the system. The Game of Life (Gardner, 1970) is the starting point of this elaborate investigation. Agents in the society share three beliefs (RED, BLUE and GREEN) with associated belief levels. The dominant belief is shared with the neighbors (if the two highest belief levels are equal, a random belief is broadcasted to the neighbors). While it would be interesting to observe how political beliefs may spread in a society, this study’s also emphasizes the sensitivity of simulations to decisions on temporality of the programmers.

The dynamics in social networks is the focus of Palla et al. (Palla, Viscek, & Barabási, 2009), in their paper titled “Statistical Properties of Community Dynamics in Large Social Networks.” It investigates social networks, as highly connected circles of friends, colleagues, or family members that grow, shrink, merge, split, appear, or disappear over time. Statistical investigations have been made possible by the explosion of the social networks online, and the databases of captured data that enable their study. This is a move away from the earlier quantitative studies that were usually based on qualitative data analysis from typically about a dozen interviewees. The strengths of the links in the network can be objectively measured via aggregating the number of e-mails, phone calls, post etc between two individuals in a network. This study uses the Clique Percolation Method (CPM) (Palla, Derényi, Farkas, & Vicsek, 2005) to study the dynamics of communities. Results indicate a significant difference between smaller circles and institutions. Small communities appear to be stable if a few strong relationships persist, which does not hold for institutions. All members of the institution might change and the institution may still exist. The authors show that the knowledge of the time commitment to a community can predict its lifetime. As such, it gives a new insight in the differences between small and large multiagent systems.

In his chapter titled “Two Informal Complexity Measures in Social Networks and Agent Communities,” Fonseca looks at the network dynamics, (Fonseca, 2009) but from the Information Theory perspective. He looks into measures: Entropy Density and Excess Entropy over the stochastic relational changing. In practice estimates can only be obtained, but nevertheless, they can be used to draw conclusions. The mechanics of the measure computation is illustrated in an artificial social network of agents where there is one coordinator for each task, an agent performs a part of the task and there is a directory facilitator. The coordinator communicates with the agents contributing to the that it is responsible for at fixed intervals, the agent communicates with the coordinator to adjust its goal, and the directory facilitator randomly exchanges messages with very agent at the agents’ request in order to facilitate the communication in the system. Fonseca studies the agent’s social predictability and the commitment to the community as characteristics of the society members. 

Neumann entropy was defined in von Neumann’s work on quantum mechanics (Von Neumann, 1955), and today it is an important tool in quantum Information Theory. In their chapter “Quantifying Complexity in Networks: The von Neumann Entropy,”  (Passerini & Severini, 2009), the authors study the Von Neumann entropy of graphs, thus transcending the fields of quantum Information Theory and Complex Networks. This marriage happens at a statistical level. Two theorems are stated and proven, and a wealth of directions for further work stated. The article proposes a new way of quantifying disorder/entropy in social networks. The results indicate that Von Neumann entropy increases with the regularity properties of the network and with the number of its connected components.

In this section we ordered the chapters with respect to the order of the emergencies they treat, from first-order to second-order emergencies. (Trajkovski & Collins, Towards More Lively Machines, 2009) This section is a collection of models that attempt to solve agency problems, and come from researchers with a variety of different backgrounds. Some attempt to solve problems that traditional Artificial Intelligence (AI) had left open; others look at societies in a brand new light.

Belief-Desire-Intention (BDI) agents (Bratman, Israel, & Pollack, 1988) are artificial agents that are capable of simple adaptations to behaviors, encoded in their hierarchical plans. These plans are indexed by goals. Execution relies on the context sensitive expansion of goals, and therefore choices and planning happens on multiple levels of abstractions. The choices the agent makes are in direct response to the current status of the environment; should the effort fail due to changes in the environment, a backtracking mechanisms is put in action. In their chapter “Enhancing the Adaptation of BDI Agents Using Learning Techniques,” Airiau et al. (Airiau, Padgham, Sardina, & Sen, 2009) outline standard  BDI approach in programming of agents, and propose enhancements to account for some of its shortfalls. The major drawback of the BDI (and really most of the traditional AI approaches) is that the coder would need to spoon-feed the machine with pairs of plans and contexts for those plans. As a proposed solution to overcome this issue the authors propose looking into the agent’s history to improve the context conditions for the plans as the agents sojourns in the given environment. The open questions are related to figuring out what to do when a plan does not exist? The authors performs experiments in so called non-deterministic environment, but then what would happen if the agent experiences perceptual aliasing, when two locally distinct places in the environment look the same to the agent? This approach is a hybrid of merging the rather typical efforts to resolve the challenges of the traditional AI approaches to agency problems and the post-AI Agents Science.

The practicality of using Reinforcement Learning (RL)  in agents presents us with the challenge of computationally heavy functions that the agent need to manipulate with quickly. The more complex the environment is to be learned is, the training set gets larger. In the training process, a value function is fitted, that serves as a measure of “how good” it is for the agent to execute a particular action while in a given state. From an engineering perspective, approximations are the next best thing to using the actual value functions, in finding a good candidate faster. The approximations are finer or coarser (or more or less expressive) based on the size of the tiles (or how fine the tessellation is) of the rank of the function. In their paper titled “Reinforcement Learning with Reward Shaping and Mixed resolution Function Approximation,” (Grzec & Kudenko, Reinforcement Learning with Reward Shaping and Mixed resolution Fubction Approximation, 2009) Grzec and Kudenko attempt to propose a hybrid approach that uses approximations to provide useful guidance of the agent, and then towards the end a more expressive approximation to get a high-quality final result. Three approaches are investigated: a combination of less and more expressive representations of one value function representation, use a less expressive function to learn the potential for reward shaping, and then use it to shape the reward of learning with a desired resolution, and a hybrid approach of the two – use less expressive approximation for the learning potential, and using it to guide learning, which combines a less and more expressive function approximation at the ground level. The paper is rich in experimental results from a mountain car (Sutton & Barto, 1998), car parking (Cichosz, 1996) and boat navigation (Jouffle, 1998) from one to another bank of the river, that are used for benchmarking purposes, and expose the mixed method as superior.

 Effective open-ended learning is what cognitive robotics strives to achieve. Open-ended learning is not task specific, and should preferably be unsupervised, which makes it inherently inductive in nature. The chapter  “Inductive Logic Programming and Embodied Agents” by Kulakov et al. (Kulakov, Lukkanen, Mustafa, & Stojanov, 2009) overviews innovative attempts to use Inductive Logic Programming (ILP)  as a machine learning technique in the context of embodied autonomous agents, and highlights novel efforts in bridging the problems between the large datasets and ILP in these agents. Artificial curiosity mechanisms drive the agents to learn. The paper overviews ILP, and explores its potential use in robotics by proposing solutions to overcome some of the ILP shortcomings.  Four general issues are discussed: overcoming difficulties by oversimplifying the environment, dealing with uncertainty, dealing with high dimensionality and noise, and discussing the scalability issue. While the merger of cognitive robotics with ILP may seem promising, some questions still remain open. Further directions of the investigation are seen in infusing probabilistic tools, layers of abstraction and work on the issue of scalability, as the environmental stimuli generates overwhelming amount of data that needs managed. Innovative heuristics may also be needed.

In the next chapter titled “Cognitive Robotics and Multiagency in a Fuzzy Modeling Framework” (Trajkovski, Stojanov, Collins, Eidelman, Harman, & Vincenti, 2009) we present a general fuzzy-logic based framework for cognitive robotics modeling. The modeling uses lattice, poset and relational structured valued fuzzy relational and algebraic structures that are applied in various capacities in cognitive agent. Five related but distinct case studies in agency and multiagency are presented. The agent(s) is (are) brought in an unknown environment and motivated to explore using inborn schemes, which represent a sequence of actions they can perform. Based on the success of an action, and the perception of the environment while the action is performed, the agent builds perception-action pairs that are  the inner representation of the environment that the agent uses to diminish pain in the new environment, by using what it had learned to anticipate the success of its next move towards getting to the place that can satisfy its active drive(s).The agent Petitagé is the base of all studies, and multiple metrics on its behavior in an unknown environment have been gathered via a simulation in the PYRO (Python Robotics) simulation environment for a single agent and in a multiagent setting. ANNA (Artificial Neural Network Agent) learns the environment using neural networks, and performs comparably to Petitagé. In order to be able to calibrate the simulations on robotic agents, data from human subjects have been harvested using POPSICLE (Patterns in Orientation: Pattern-Aided Simulation Interactive Context Learning Experiment), and the Izbushka experimental setup. 

To calibrate simulations of multiagent environments it is necessary to be able to have enough data from the agents that are being modeled. The Izbushka experiment is discussed in detail in the chapter “Designing Izbushka: Investigating Interactions in Context Zero Environments” (Braman, 2009). It investigates human behavior in context of varying side. This experiments aims to understand how human agents learn in a Petitagé-like environment. Izbushka is an agent in itself that monitors the success of the human that interacts with it, and based on the observations presents the human subject with the environment, which is dynamically changing. Both qualitative and quantitative data is collected from the subjects and analyzed. As such Izbushka, even on its own is a bona fide autonomous intelligent agent that couples with the human agent thus forming a heterogeneous multiagent system. The repercussions of this study are far-reaching in the efforts to achieve true virtual personal assistants that learn from and about the user, which is basically the underlying idea of the Web 3.0 hype and hope.

Through interaction in multiagent societies, societal norms emerge and propagate between the population. In the essay on emergence of social norms, “Norm Emergence with Biased Agents,” Mukherjee et al. (Mukherjee, Sen, & Airiau, 2009) observe phenomena in a multi-agent simulation where the agent are randomly paired and interact with each other privately. Each agent is learning from the agents they are paired with over time. The simulations in the study were carried in three types of populations: with same initial bias, in a population where initially 50% of the population shares a bias that is opposite than the initial bias of the other half, and  an initial population where the opposing norms are unevenly distributed. The society is homogenous and uses the WoLF-PHC (Win or Learn Fast – Policy Hill Climbing) algorithm for learning norms. (Bowling & Veloso, 2002)   While this algorithm converges in a two person, 2 actions game against an opponent in a Nash equilibrium, it is not known if that would be the case with social learning. The experiment is modeled after the game of traffic where drivers arrive at intersections simultaneously. In these societies the norms emerge via a bottom-up process that depends on direct individual experiences, and not on gossip or observations. These interactions happen with random strangers in the society that were originally biased. One notable observation is that norms converge faster when there the majority is larger. This study opens the doors for many related simulations with varying parameters, such as the population size, varying population size, number of original biases, expanding the social learning to include observations and gossip, and many more.

A case study in norm dynamics is the presented in the following chapter. Distributed online communities have been experiencing wide acceptance and use. Goldspink’s paper “Social Self-Regulation in Computer Mediated Communities: The Case of Wikipedia” (Goldspink, 2009) focuses on the Wikipedia community, and investigates how social self-regulation is achieved via normative mechanisms. It is an empirical study of norm innovation in this system. Although novel users are typically surprised when they encounter Wikipedia, and may expect that such an open collaborative system is not likely to produce credible encyclopedic entries, reality check witnesses to the contrary, despite the nonexistence of the typical editorial checks and balances involved in a classical encyclopedic project. (Giles, 2005) The question examined here is how social communication influences the behaviors of others in open systems, and particularly the processes that enable self-regulation, whether the findings are consistent with existing theories, and what alternative hypotheses can be drawn based on this case study. In Wikipedia, much alike any other wiki, there are two main activities: editing and conversations about editing. This article studied discussion pages on selected controversial articles in Wikipedia. Although the norms that emerged could be seen all over the place, in Wikipedia order appears to have emerged. Due to the open contribution and the participation of anonymous volunteers, the neutral style of communication is prevalent. The communication style and emerged norm do not appear to have a lot to do with the content of the articles being edited. The paper states other hypotheses as well that challenge some existing theories. Although this institution is based on voluntary participation, and there is no product involved, as in e-commerce sites, for example, it represents a solid kick-off to a sizable study of social norm innovation in open institutions.

Multiagent technologies and systems provide a range of tools that are increasingly being used to solve real-world problems. This section consists of a collection of chapters that emphasize how models and approaches from this discipline can give valuable insight into challenges that range from energy consumptions and crises in the stock markets, to distance education.

The energy utilization system is a dynamic one, and experiences steep ups and downs in usage in the traditional manner. In their paper “A Reinforcement Learning Approach to Setting Multi-Objective Goals for Energy Demand Management,” Guo et al. (Guo, Zeman, & Li, 2009) discuss an agent technology application for energy demand management, where appliances are individual agents that can defer their energy consumption if there is a need on the power system for them so to do. High demand of electricity equals higher price for this utility. When the energy demand is high, agents collaborate in a multiagent environment and defer the energy consumptions by appliances when the energy price peaks. The study simulates the Australian power brokerage system. It highest priority is the stability of the power system, so the system has two goals – maintain that stability and optimizing the cost. A cap function is defined to denote the capped energy consumption, and its behavior observed in a multi-agent simulation. The energy broker is dually rewarded (using the two goals) to accommodate the attentive plans for energy consumptions of the individual agents. Two simulations are presented, one featuring the feedback of the group agents (aggregation of feedback from individual agents/appliances), and one without, showing consistent improvement when the two goals are used in the reward functions. It turns out that price itself is not a reward function enough to keep this system stable, and avoid peaks in usage of power. 

In their paper “A Step-by-Step Implementation of a Hybrid USD/JPY Trading Agent,” Barbosa and Belo (Barbosa & Belo, 2009) present the details of a successful software agent that trades US and Japanese currency (denoted by USD and JPY respectively) on the Foreign Exchange Market (Forex). The agent uses prediction and mitigates risk. Taking into account the broking fees, the system over 17 months generated a return of about 50%. The system is based on simple trending principles that have proved successful in the trading of the USD/JPY currency pair in a 6 hours timeframe. This is an encouraging agent, but as the authors indicate, it would be a mistake to use it life just yet on the stock exchange, as a successful trading is based on trading a diversified portfolio. And, for a diversified trading activities, more and different other agents - implementing a diversified range of trading techniques - will need to be developed.

Araúho and Louça, in the paper titled “Modeling Multi-Agent Systems as a Network: A Methaphoric Exploration of the Unexpected,” explore the S&P500 stock market over the years. (Araújo & Louça, 2009) The S&P 500 index have been monitoring 5000 large-cap common stocks in the United States since 1957. Using mathematical approaches, namely geometrical and topological characterizations (with the appropriate metrics), the paper shows how the stocks have been clustering and re-clustering over the years, forming clusters in the network of stocks. Case studies of “rough” and “smooth” periods of the market are shown. The authors construct a coefficient that quantifies the distribution and intensity of the correlations among stocks in the market over a decade. This coefficient clearly marks the crises on the market in 2001-2002, and in 2008.

Romero (Romero, 2009), in the chapter “Virtual Worlds and the Implication for Accountants: The Case of Second Life,” discusses the emerging economies of the virtual world of Second Life. As its popularity is growing, so are its real world financial implications, as members spend a lot on virtual items and virtual real estate transactions, and some residents generate hefty incomes in real world dollars. Second Life (Linden Research) is still in an early stage of development, and there are many financial and legal regulatory issues to be resolved.  But with these challenges come opportunities; Second Life may be the impetus for a new accounting platform that may bring different practices together and provide new growth opportunities that financial communities have been looking for in the virtual world. The emerging economies are phenomena worth monitoring in this real-world online multiagent system.

CIUCEU (Comunicação Infrormal entre Utilizadores de Correio Electrónico Universitário – Informal Communication between Users of a University E-mail System) is a model developed using MASON (Multi-Agent Simulator of Neighborhoods), (Luke, Balan, Sullivan, Panait, Cioffi-Revilla, & Paus)  and is the focal point of the Rodrigues’ research presented in the chapter  titled "CIUCEU: Multi-Agent-Based Simulation of University Email Communities." (Rodrigues, 2009) It models the informal communication via e-mail at his affiliate institution. The data from communication between teachers, students, and employees of the institution, over a short period of 62 days was observed. The simulation shows that even this period provided sufficient data to capture the latent structure of the network. The observation of the clique percolation, for example, reveals that vertical hierarchies are very loose and communication transcends departments. Cluster algorithms have been applied to gather conclusions on the groupings of the users. The research also emphasizes the need for careful consideration of the setup of the simulation, and making a balance between a totally open system and some rules (introduced from the real data) in order to gather insights that would be interpretable. 

Welzant’s  study of a specific student experience in an online system of distance education, titled “Impact on Learner Experience: A Qualitative Case Study Exploring Online MBA Problem-Based Learning Courses,” (Welzant, 2009) concludes that there is a significant need for problem-based learning (PBL) in online courseware. In agent terms, students prefer learning based on new environments and via exploration. The various communication methods used in the distance education arena today enable agent interaction and collaboration when solving problems they are faced with PBL challenges.

Ho et al., (Ho, Enz, Dautenthahn, Zoll, Lim, & Watson, 2009), in their chapter titled “Towards Learning ‘Self’ and Emotional Knowledge,” present research towards the development of a virtual learning environment inhabited by intelligent virtual agents and modeling a scenario of inter-cultural interactions. This is a part of the research of a team focused on emotional knowledge learning in autobiographic social agents, and aims to promote intercultural empathy. The ultimate aim of this environment is to allow users to reflect upon and learn about intercultural communication and collaboration. Rather than predefining the interactions among the virtual agents and scripting the possible interactions afforded by this environment, the authors pursue a bottom-up approach whereby inter-cultural communication emerges from interactions with and among autonomous agents.  The intelligent virtual agents that inhabit this environment are expected to be able to broaden their knowledge about the world and other agents, which may be of different cultural backgrounds, through interactions. 

Works Cited
Airiau, S., Padgham, L., Sardina, S., & Sen, S. (2009). Enchancing the Adaptation of BDI Agents Using Learning Techniques. (G. Trajkovski, Ed.) IJATS , 1 (2), 1-18.

Araújo, T., & Louça, F. (2009). Modeling Multi-Agent Systems as a Network: A Methaphoric Exploration of teh Unexpected. (G. Trajkovski, Ed.) IJATS , 1 (4), 17-29.

Barbosa, R. P., & Belo, O. (2009). A Step-by-Step Implementation of a Hybrid USD/JPY Trading Agent. (G. Trajkovski, Ed.) IJATS , 1 (2), 19-35.

Bowling, M. H., & Veloso, M. M. (2002). Multiagent Learning Using a Variable Learning Rate. Artificial Intelligence , 136 (2), 215-250.

Bragin, J. (2009, October). Review of the "Handbook of Research on Agent -Based Societies: Social and Cultural Interactions". Retrieved June 6, 2010, from Journal of Artificial Societies and Social Simulation:

Braman, J. (2009). Designing Izbushka: Investigating Interactions in Context Zero Environments. IJATS , 1 (1), 74-95.

Bratman, M. E., Israel, D. J., & Pollack, M. E. (1988). Plans and Resource-Bounded Practical Reasoning. Computational Intelligence , 4 (4), 349-355.

Cichosz, P. (1996). Truncated tempporal Diffeernces with Function Approximations: Successful Examples using CMAC. Proceedings of the 13th European Symposium on Cybernetics and Systems research. 

Conover, A. J., & Hammell, R. J. (2009). Agent Interaction via Message-Based Belief Communication. (G. Trajkovski, Ed.) IJATS , 1 (3), 1-28.

Conover, A. J., & Trajkovski, G. (2007). Effects of temporary asynchronous interaction on simple multiagent behavior. Emergent agents and socialities; Social and organizational aspects of intelligence. fs-07-04, pp. 34-41. Menlo Park: The American Association for Artificial Intelligence.

Erdi, P. (2008). Complexity Explained. Berlin & Heidelberg: Springer-Verlag.

Fonseca, A. J. (2009). Two Informal Complexity Measures in Social Networks and Agent Communities. (G. Trajkovski, Ed.) IJATS , 1 (4), 58-67.

Frasier, J. (1998). From Chaos to Conflict. In J. Frasier, M. P. Soulsby, & A. Argyros, Time, Order, Chaos. Madison, Conn: International University Press.

Gardner, M. (1970, October). MATHEMATICAL GAMES: The fantastic combinations of John Conway's new solitaire game "life". Scientific American , pp. 120-123.

Gilbert, N. (2008). Agent-Based Models. London: SAGE Publications.

Giles, J. (2005, February 2). Internet encyclopaedias go head to head. Nature , 438 (7070), p. 890.

Goldspink, C. (2009). Social Self-Regulation in Computer Mediated Communities: The Case of Wikipedia. (G. Trajkovski, Ed.) IJATS , 1 (1), 19-33.

Grzec, M., & Kudenko, D. (2009). Reinforcement Learning with Reward Shaping and Mixed resolution Fubction Approximation. (G. Trajkovski, Ed.) IJATS , 1 (2), 36-54.

Grzec, M., & Kudenko, D. (2009). Reinforcement Learning with Reward Shaping and Mixed resolution Function Approximation. (G. Trajkovski, Ed.) IJATS , 1 (2), 36-54.

Guo, Y., Zeman, A., & Li, R. (2009). A reinforcement learning Approach to Setting Multi-Objective Goals for Energy Demand Management. (G. Trajkovski, Ed.) IJATS , 1 (2), 55-70.

Ho, W. C., Enz, S., Dautenthahn, K., Zoll, C., Lim, M., & Watson, S. (2009). Towards Learning ‘Self’ and Emotional Knowledge. (G. Trajkovski, Ed.) IJATS , 1 (3), 51-78.

Jouffle, L. (1998). Fuzzy Inference System Learning by Reinforcement learning. IEE Transactions of Systems, Man, and Cybernethics, Part C: Applications and Reviews , 28 (3), 338-355.

Koehler, G. (2009). Attending to Temporal Assumptions May Enrich Autonomous Agent Computer Simulations. (G. Trajkovski, Ed.) IJATS , 1 (1), 1-18.

Kulakov, A., Lukkanen, J., Mustafa, B., & Stojanov, G. (2009). Inductive Logic Programming and Embodied Agents. (G. Trajkovski, Ed.) IJATS , 1 (1), 34-49.

Linden Research. (n.d.). What is Second Life. Retrieved June 5, 2010, from Second Life:

Luke, S., Balan, G. C., Sullivan, K., Panait, L., Cioffi-Revilla, C., & Paus, S. (n.d.). MASON. Retrieved June 2, 2010, from

Macy, M. W., & Willer, R. (2002). From Factors to Actors: Computational Sociology and Agent-Based Modeling. Annual Review of Sociology , 28, 143-166.

Mukherjee, P., Sen, S., & Airiau, S. (2009). Norm Emergence with Biased Agents. (G. Trajkovski, Ed.) IJATS , 1 (2), 71-83.

Palla, G., Derényi, I., Farkas, I., & Vicsek, T. (2005). Uncovering the Overlapping Community Structure of Complex Networks in Nature and Society. Nature , 435, 814-818.

Palla, G., Viscek, T., & Barabási, A.-L. (2009). Statistical Properties of Community Dynamics in Large Social Networks. (G. Trajkovski, Ed.) IJATS , 1 (4), 1-16.

Passerini, F., & Severini, S. (2009). Quantifying Complexity in Networks: The von Neumann Entropy. (G. Trajkovski, Ed.) IJATS , 1 (4), 58-67.

Rodrigues, D. (2009). CIUCEU: Multi-Agent-Based Simulation of University Email Communities. (G. Trajkovski, Ed.) IJATS , 1 (4), 30-48.

Romero, J. A. (2009). Virtual Worlds and the Implication for Accountants: The Case of Second Life. (G. Trajkovski, Ed.) IJATS , 1 (3), 45-50.

Sutton, R. S., & Barto, A. G. (1998). Reinforcement Learning: An Introduction. Cambridhe: MIT Press.

Trajkovski, G. (Ed.). (2009). International Journal of Agent Technologies and Systems , 1 . Hershey, PA, USA: IGI Publishing.

Trajkovski, G., & Collins, S. G. (Eds.). (2009). Handbook of Research on Agent-Based Societies: Social and Cultural Interactions. Hershey, PA, USA: Information Science Reference.

Trajkovski, G., & Collins, S. G. (2009). Towards More Lively Machines. In G. Trajkovski, & S. G. Collins (Eds.), Handbook of Research on Agent-Based Societies: Social and Cultural Interactions (pp. xvii-xxii). Hershey, PA , USA: Information Science reference.

Trajkovski, G., Stojanov, G., Collins, S., Eidelman, V., Harman, C., & Vincenti, G. (2009). Cognitive Robotics and Multiagency in a Fuzzy Modelling Framework. (G. Trajkovski, Ed.) IJATS , 1 (1), 50-73.

Von Neumann, J. (1955). Mathematische Grundlagen der Quantenmechanik (Mathematical Foundations of Quantum Mechanics). Berlin: Springer.

Welzant, J. H. (2009). Impact on learner experience: a qualitative case study exploring online MBA problem-based learning courses. (G. Trajkovski, Ed.) IJATS , 1 (3), 29-44.

Wilensky, U. (2009, December 20). NetLogo 4.1 User Manual. Retrieved June 2, 2010, from NetLogo: 

Author(s)/Editor(s) Biography

Dr. Goran Trajkovski is the founding Editor-in-chief of the International Journal of Agent Technologies and Systems. He is a partner and the CEO of Algoco eLearning Consulting, and the VP for Business Development for the Americas at Kreofina. In the past, he has served as faculty at different institutions, as well as the Director of Product Strategy and Development at Laureate Education, Inc., and the Chair of the Department of Information Technologies at the Education Management Corporation (EDMC)/South University, USA. In 2003, he founded the Cognitive Agency and Robotics Laboratory (CARoL) at Towson University, Towson, MD, USA. Dr. Trajkovski’s research focuses on cognitive and developmental robotics, and interaction and emergent phenomena in societies of agents. He is an affiliate of the Institute for Interactivist Studies at Lehigh University, and a member of the organizing committee of the biannual Interactivist Summer Institutes. He has authored over 300 publications, including twelve books and edited volumes. He has chaired two symposia for the Association for Advancement of Artificial Intelligence. His work has been funded by NSF, the National Academies of the Sciences, and OWASP (Open Web Application Security Project). Dr Trajkovski hold a BSc in Applied Informatics, MSc in Mathematical and Computer Sciences, and PhD in Computing Sciences from the University “SS Cyril and Methodius,” Skopje, Macedonia.