Toward Agent-Based Software Engineering for Information Access Applications

Steven C. Laufmann
U S WEST Advanced Technologies
4001 Discovery Drive
Boulder, CO  80303
laufmann@advtech.uswest.com
 

Agents for Information Access

For its first few decades, computer science research and development was focused on computational problems and their solutions. The solutions were typically centered on the best possible methods for computing unambiguous mathematical problems. The solutions tended to be grounded in the hardware. These solutions were foundational, and provided the financial incentives necessary for the field to become established and grow.

However, in the last two decades there has been increasing attention on a new set of problems those dealing with the interactions between multiple computational entities. Such attention started with fairly simple interprocess communication, but has grown to include the much more difficult problems associated with coordination, cooperation, and collaboration between and among computational entities. Such problems are inherently complex, and often involve semantic ambiguities.

These problems are characteristic of a large class of applications associated with real-world information access. While many of these problems are inherently difficult, their solutions promise substantial organizational and commercial benefits.

A real-world example, shown in Figure 1, is a view of a complex order processing system, which involves a large number of humans and computing systems that must interact in various ways. The process embodied in this diagram grew over a long period of time, and essentially without global design or control. The net result is an overly complex system that consists almost entirely of complex and often redundant interactions. A very large majority of the software and the ongoing maintenance involved in the overall system is taken up with these interactions, which until recently included staffs of humans that read the output from one system and typed it in to the next. The overall system is an example of the detrimental costs associated with inadequately considering the coordination and cooperation aspects of large system development.

In this example, as in a large number of similar cases, the degree to which the coordination, cooperation, and collaboration issues are successfully managed is the degree to which the overall system is successful. It is this set of problems, and their solutions, which will shape the future of computing.
 
 

Figure 1. Order Processing System
 

The Power of Agents

The notion of agents offers an attractive alternative to the costly and inefficient approach to interactions above, by offering the opportunity to formalize and deploy adequate models of coordination, cooperation, and collagoration. These advantages come in two ways:  

The Future of Agents

Agents as the Future of Computer Science

The agent community is, then, addressing the right set of issues. If adequate models can be designed and agreed upon, and subsequently translated into software development metaphors and appropriately implemented as software development mechanisms, the potential is enormous.

However, the solutions to these problems currently lack maturity. Perhaps the best approach is to start with the least complex of the problems, the more simple forms of coordination, then work toward more complex models that address cooperation and collaboration.

Agents in the Mainstream

The ultimate goal is for the metaphor and mechanisms associated with agents to become an integral part of the computing mainstream. It seems likely that this will happen in stages:

References

"Agent Software for Near-Term Success in Distributed Applications" S. C. Laufmann. In Agent Technology: Foundations, Applications, and Markets, eds. Nicholas R. Jennings and Michael J. Wooldridge, pp. 49-69. Springer-Verlag.

"Toward Agent-Based Software Engineering for Information-Dependent Enterprise Applications" Steve Laufmann. IEE Proceedings on Software Engineering 144(1): 38-50. January, 1997.

"The Information Marketplace: Achieving Success in Commercial Applications" Steve Laufmann. In Electronic Commerce: Current Research Issues and Applications, Lecture Notes in Computer Science 1028, Eds. Nabil Adam and Yelena Yesha, pp. 115-147. Springer-Verlag.

"The Information Marketplace: The Challenge of Information Commerce" Steve Laufmann. Second International Conference on Cooperating Information Systems (CoopIS-94). May, 1994.

"An Organizational Framework for Cooperating Intelligent Information Systems" Mike P. Papazoglou, Steven C. Laufmann, and Timos K. Sellis. International Journal for Intelligent and Cooperative Information Systems 1(1):169- 202. March,1992.

"Coarse-Grained Distributed Agents for Transparent Access to Remote Information" Steven C. Laufmann. The Next Generation of Information Systems: From Data to Knowledge, Lecture Notes in AI, p. 223-237. Eds. M. Papazoglou and J. Zeleznikow. Springer-Verlag. 1992.

"Direct End-User Access to Remote Information" Steven C. Laufmann, Richard L. Blumenthal, Laural M. Thompson, and Beth Bowen. Proceedings of the 1991 Conference on Organizational Computing Systems (COCS-91). November, 1991. pp. 16-28.

"Communication and Cooperation Among Coarse-Grained Distributed Agents" Steven C. Laufmann, Mitchell J. Nathan, Richard L. Blumenthal. AAAI-91 Workshop on Cooperation Among Heterogeneous Intelligent Systems. July, 1991.
 


Position Paper. International Workshop on Agent-Oriented Information Systems, Seattle, May 1999. See http://www.AOIS.org