Agent-Oriented Information Systems Information systems are and continue to be the predominant application of computing technologies. The development, maintenance, and evolution of information systems is the primary preoccupation of many computing professionals. In spite of ongoing advances in many areas of information technology, computing professionals continue to face numerous challenges. Information systems are now placed in ever more demanding roles, in distributed and networked environments, and with ever higher expectations from users -- in functionality, as well as non-functional qualities such as robustness, responsiveness, flexibility, etc.
Agent orientation has a lot to offer to information systems, both in terms of enabling technologies and in terms of conceptualizing what information systems are and can be, and how they can be developed, sustained, and evolved.
Despite this potential, agent concepts have mostly played only a niche role in information systems, for example, in offering specialized capabilities in information searching and brokering. The Agent-Oriented Information Systems (AOIS) workshop series aims to promote and stimulate interest and discussion in an agent-oriented conception of information systems, covering all aspects of information systems, and all stages of development -- requirements analysis, design, and implementation.
To promote cross-fertilization of ideas between the AI agents community and the information systems community, the AOIS workshop was held at two locations this year, at AAAI in Austin, Texas, and at the Conference on Advanced Information Systems Engineering (CAiSE) in Stockholm, Sweden. This year's workshop continues the success of AOIS'99, which was held at the Autonomous Agents'99 conference in Barcelona and at CAiSE'99 in Heidelberg, Germany.
Approximately 45 participants attended AOIS-2000, with around 25 coming to the AAAI/Austin location of the workshop. The program in Austin included invited presentations by James Odell and Bill McCarthy, and eight paper presentations dealing with a wide range of issues in the area.
The first invited talk, by well-known object-oriented software development author James Odell, discussed efforts within FIPA and the OMG Agents Working Group to develop and standardize extensions to the UML modeling language to support the design of agent-based systems. He described several aspects of the extended language Agent UML that is being developed, with particular attention to modeling agent interaction protocols. This work is important for furthering the use of agent technology in industry by relating agent-oriented models to existing techniques.
In the other invited talk, Bill McCarthy, a well-known business and information systems researcher, presented his work on the REA (Resource-Event-Agent) model of economic transactions, and how it has been applied to represent enterprise value chains and workflows. He discussed how the model might be used to develop agent-based systems where the agents act as full-fledged economic participants.
Among the paper presentations, Cernuzzi and Giret described their work on comparing different agent-oriented software engineering methodologies and discussed a case study that was done using Kinny, Georgeff, and Rao's (KGR) methodology. The KGR methodology was criticized for its lack of support for the specification of fuzzy goals and behavior evolution, and for translating a design into an implementation.
Vercouter, Beaune, and Sayettat presented a decentralized approach to the integration of new agents in an open multiagent system. Agents learn stereotypes to classify their acquaintances. When they encounter a new agent, they interact and form a description of it, classify it, and then can recommend acquaintances to it.
Camacho, Molina, and Borrajo described their experience in building a multiagent system for electronic travel planning using Web information resources. The architecture involves planning agents based on the Prodigy planner and Web agents that access information sources and interpret the results.
Filipe argued that socially-aware agents need to be able to cope with different norm systems. He presented a model inspired from the Information Field paradigm for how agents might resolve normative conflicts based on their values and beliefs.
Zhu, Greenwood, Huo, and Zhang discussed how quality management in information system was being affected by agent-orientation, the Web, and other recent changes in information technology. They presented a model of information system evolution through "growing up" and sketched how agent-based quality management tools might be designed to cope with this.
Vinaja, Slinkman, and Mykytyn presented the results of an experimental study of the effects of internet information delivery and agent facilitation on a decision making task. The study found that agent facilitation leads to fewer information sources being examined and higher satisfaction with the process. Workshop participants were interested by the approach but questioned whether the results could be generalized beyond the system studied.
Kaminka, Pynadath, and Tambe presented a non-intrusive approach to the monitoring of complex multiagent systems based on eavesdropping on the messages exchanged by the agents. The approach performs plan recognition on the messages and uses a model of team coherence to infer the system state more accurately.
Mukherjee, Dutta, and Sen discussed the use of agents to support query reformulation, a key problem in information retrieval. Such agents will need rich domain ontologies to perform their task. Characteristics of domain ontologies that facilitate the reformulation task are identified, in particular the presence of suitable ordering relations.
Among the wealth of impressions and questions that were left in mind from the workshop, there are two that perhaps stand out. One is the enormous breadth of the AOIS domain. Our goal is to build systems modeled after social organizations, that will be deployed in the real world, with all the engineering skill that this requires. For this, we will need the contributions of people from many disciplines. The second is the importance of reaching out to practitioners, of presenting agent-based concepts and methods in ways that ease the transition from traditional approaches.
The number of submissions and participants, and the interactions at the workshop suggest that AOIS will develop into an important area both from the AI agents standpoint and from the information systems standpoint. The organizers are planning to continue holding the workshop to facilitate and promote the development of this area. Further information about this workshop and future ones can be found at http://aois.org. -