Introduction

Today’s databases are increasingly inter-connected; desktop workstations and computers are increasingly inter-connected; and the software interfaces which provide entry into the data environments are intuitive and accessible.

What we called "information overload" in the past has become a deluge, the celebrated "firehose" of the Internet. With the penetration of client/server architectures, online databases, Internet connectivity, electronic mail and online services, too many people have access to far too much information to digest and utilize in an intelligent fashion. Software filters, culling unwanted information has helped but do not go far enough. Another, more powerful solution is called for. That solution is intelligent agent software, software which actively pursues information appropriate to the users needs while they go about their business.

Intelligent agents were impractical until recently. To be of use, agents require a constant flow of data in large volumes. Internet connectivity provides that flow. Computing hardware had to achieve a level of speed and robustness, on a common scale to ensure that the processing of data, and the necessary algorithms and calculations could be accomplished in a reasonable time. In addition, software agents must be active behind interfaces immediately accessible and intuitive. The Web browser provides this access. The inter-connectivity of computers and people, the development of intuitive software interfaces, and the wide spread implementation of each of these in the past two years has set the stage for the development of agent software.

In addition to measuring the relationships between data and user, the agent will measure the quality or granularity of retrieval success rates. The agent measures the relationships between the user profile, the available data attributes, user action on the available data, and user action based on agent recommendations. Calculations on all of these help the agent learn and modify its behavior when necessary, even prompting the user to alter the profile when information retrieval is consistently distinct from the stated profile. Repeated statistical analysis of information events ensure continued accuracy. As this accuracy is sustained, an increasingly refined review and presentation of information is possible.