Gerd Wagner (Univ. of Leipzig) and Eric Yu (Univ. of Toronto)
The First International Workshop on Agent-Oriented Information Systems (AOIS'99) was held as a bi-continental bi-conference event in Seattle (USA), 1 May 1999, in conjunction with Agents'99, and in Heidelberg (Germany), 14-15 June 1999, in conjunction with CAiSE*99. The workshop consisted of 5 invited talks, 2 panels, and 9 refereed paper presentations selected from 19 submissions from 11 countries. All papers and the slides of the invited talks can be obtained from http://www.AOIS.org/99.
To foster greater communication and interaction between the Information Systems and Agents communities, the AOIS workshop is organized as a bi-conference event intended to be a single "logical" event with two "physical" venues. It is hoped that this arrangement will encourage greater participation from, and more exchange between, both communities. While only a small number of people was actually able to attend both venues of the workshop, we hope that in the forthcoming years more and more people will cross the bridge between the two communities and their respective research agendas, created through the bi-conference format, in both directions.
In organizing this workshop, we aimed to bring together researchers and practitioners in the agents area and in the information systems community to discuss the increasing role that agent concepts and techniques are beginning to play in information systems and information systems engineering. A primary objective of AOIS'99 was therefore to help define this new area of agent orientation in information systems. In this regard, we believe this inaugural event has succeeded in laying the groundwork for a promising research area. We thank and congratulate all the authors, speakers, panelists, participants, and last but not least, all the program committee members for having made the workshop a success.
The invited talks gave a variety of complementary views on agent-orientation in information systems. Katia Sycara presented the notion of middle agents and the agent capability description language LARKS used by middle agents to pair service-requesting agents with service-providing agents that meet the requesting agents' requirements. The LARKS matchmaking process employs techniques from information retrieval, AI, and software engineering to compute the syntactical and semantic similarity among agent capability descriptions for enabling heterogeneous information systems to interoperate. Interaction-oriented programming, as conceived by Mike Huhns, is based on a concept of mental state that comprises social commitments and team intentions as central components. As a general approach to software engineering, interaction-oriented programming can also be used for building distributed information systems consisting of a variety of agents such as user agents, middleware agents, and information resource agents that access, e.g., databases and legacy systems.
John Mylopoulos stressed the importance of agent-oriented concepts in early requirements analysis. He pointed out that, traditionally, IS engineering has been implementation-driven. Agent-oriented concepts and methods, now, may open the door to requirements-driven IS engineering where the analysis and design methodology is no longer dictated and compromised by the prevailing implementation techniques. An important issue in any open environment is security and trust. Cristiano Castelfranchi pointed out that not only may there be malicious agents that lie and steal, but software agents and information systems may also have good reasons to deceive us and other software agents, e.g. in order to get a good deal in bargaining, or to protect confidential information from unauthorized access. Thus, a theory of trust and deception is needed in order to establish criteria and mechanisms for trust building.
It is interesting to note that the talks of Sycara and Huhns focused on software agent technologies for innovative information system, whereas Mylopoulos and Castelfranchi emphasized the need to start from the social nature of natural agents, or actors. The ensuing discussions made clear that agent-orientation in information systems includes the treatment of both artificial and natural agents, as well as their complex social interactions. Software agent technologies will help to realize advanced IS functionalities and architectures, while the consideration of extra- and intra-organizational actors as first-class citizens in the analysis and design of IS will lead to more effective requirements analysis, system design and implementation.
Finally, Stefan Kirn, in his invited talk, presented a list of questions that have to be addressed in order to successfully introduce agent-orientation in real-world business settings: How to incorporate agents into large business applications? How to benchmark agent architectures? Does agent technology meet industrial standards? Is there a competitive advantage through the use of agent technology? A new research program, Intelligent Agents and Realistic Business Scenarios, that has been recently approved by the German Research Foundation to start in the year 2000, will hopefully answer some of these questions.
Three of the contributed talks were concerned with agent-oriented modeling. Mohamed Elammari presented a methodology on how to model agent-based systems using use case maps. Lei Yu proposed a conceptual framework for agent-oriented workflow modeling. Interestingly, both papers consider deontic state components, such as authorizations, obligations and permissions, as essential attributes of agents. Kuldar Taveter presented an agent-oriented approach to business process modeling where he relates business rules to beliefs, goals and communication acts.
Two contributions presented agent-oriented approaches to particular
types of information systems. Jian-Yun Nie described an agent-based distributed
architecture for digital library systems consisting of multiple
databases and a number of search, presentation and query agents.
Martin Kollingbaum proposed to integrate both agent and database technologies
in order to represent products in a manufacturing control system
as persistent agents that are connected to a physical manufacturing process
via sensor messages.
Four papers dealt with agent-oriented software engineering issues.
Ralf Kühnel presented an assistant agent model for assisting
in finding and using certain file processing services such as printing,
converting or extracting. The model includes a planning facility for configuring
composed tasks. It has been implemented with Java technology. Khaled Nagi
proposed to use the database transaction concept for handling concurrent
actions of agents. Christopher Landauer proposed a knowledge-based approach
to an integration infrastructure that has been developed over the last
decade. Scott DeLoach described a comprehensive multiagent system engineering
methodology that supports the use of formal methods.
The panels added much to the lively discussions at the workshop. Panelists at the Seattle workshop included Frank Dignum of the Technical University of Eindhoven, Mike Huhns, and Serge Mankovski of Mitel Corporation. The panel at the Heidelberg workshop consisted of C. Castelfranchi, M. Huhns, S. Kirn, and J. Mylopoulos. The workshop was attended by about 25 people at the Seattle location and about 20 in Heidelberg.
A special highlight of the AOIS'99 workshop was the dinner at the historical
Heidelberg restaurant "Zum Güldenen Schaf". Inspite of the miserable
translation of the menu by one of the workshop organizers, all participants
managed to get a tasty meal. Finally, the owner of the restaurant
surprised the workshop participants with a performance on his hurdy-gurdy
and with a guided tour of the spectacular restaurant museum, thereby increasing
customer satisfaction and return on investment of all participants.
It is hoped that AOIS'2000 will maintain the high standards set by this
event.