The research and application area of intelligent information agents is of rapidly increasing importance. Information agents are
computational software systems that have access to multiple, heterogeneous and geographically distributed information
sources. One of their main tasks is to perform active searches for relevant information in non-local domains on behalf of their
users or other agents. This includes
retrieving, analyzing, manipulating,
and integrating information available from multiple
autonomous information sources. The information agents have to face up to the increasing complexity of modern information
environments, ranging from relatively simple in-house information systems, through large-scale multidatabase systems, to
the visionary Infosphere ('Cyberspace') in the Internet. These environments are
open and can dynamically change over time.
To cope with such information environments means, in particular, to deal with uncertain, incomplete and vague information.
Moreover, the need for
human-agent interaction in such environments, for example via synthetic characters, believable
avatars or multi-media-based representation of the partly 'fuzzy' information space available for individual users on the
Internet, remains a challenging research topic. The effective handling of uncertainty is critical in designing, understanding,
and evaluating computational systems tasked with making intelligent decisions.
In general, it is particularly important to investigate to what extent
methods from
Artificial Intelligence, Database Systems
and Information Retrieval can be applied to information discovery by single or groups/teams of information agents in the
Internet and the World Wide Web. This concerns, e.g., the use of methods for
interoperability among information systems of
all kinds, efficient techniques from machine learning, evolutionary computing, and symbolic/numerical approaches for
uncertainty reasoning as well as information retrieval in sources with semi-structured or multimedia data. Moreover,
commercial aspects of information gathering on the Internet are becoming more and more relevant: for example, agents are
paid and have to pay for services (
Electronic Commerce and Virtual Agent Marketplaces). Thus, methods for rational,
utility-based cooperation among agents are needed. In addition, mobile information agents seem an attractive means of
providing flexible and efficient information discovery in constrained environments.
Information agents may be classified into:
- Cooperative and non-cooperative information agents
- Adaptive information agents, e.g. personal assistants for information searches on the Web or collaborating information
agents that adapt themselves as a system in changing environments.
- Rational information agents that, in an economic sense, behave in a rational or utilitarian manner, and might form
temporary coalitions to increase their individual or collective
benefits
- Mobile information agents that can faciliate, among other things, dynamic load balancing in large-scale networks,
migration of business logic within medium-range corporate intranets on demand, and the reduction of data transfer
among information servers and applications.