AGS Abstract

Agent Services

A dissertation by José Alfredo Sánchez,
B.S., Universidad de las Américas--Puebla, México;
M.S., Texas A&M University
Chair of Advisory Committee: Dr. John J. Leggett
(August 1996)

ABSTRACT

Agents recently have gained enormous popularity both in computer science research and in the software industry. Although the research projects addressing agency issues are legion, a precise and generally accepted definition of agent has yet to emerge. Based on an extensive survey of the field, this dissertation presents a taxonomy of agents that provides a high-level perspective of the existing views on agency and their inter-relationships. The three classes of agents on the first level of this taxonomy, each corresponding to a major view, are programmer agents, network agents and user agents. The motivations, characteristics, applications and underlying theories for each class are discussed. The class of user agents, which constitutes the main theme of the dissertation, is further elaborated. User agent subclasses and a framework for characterization of commonly used techniques employed in user agent implementation are presented.

Agents that perform tasks on behalf of the user have great potential for assisting users in coping with large, complex, and dynamic information spaces such as digital libraries. This dissertation also presents AGS, an open architecture designed to provide agent services to participants of digital libraries. In AGS, specific user agent functionality is defined collaboratively by librarians, who are experts in the organization of digital library resources, and subscribed end users (or digital library patrons), who determine what library resources are of interest. Users delegate tasks to, visualize, and control agents in multiple libraries via a uniform interface. The design of AGS strongly fosters agent control by the user, provides extensible agent functionality, and utilizes readily available technologies in the construction of an environment that actively cooperates with the user in taking advantage of information resources. AGS enhances current views on services to be enabled by digital libraries and offers a much needed testbed for the investigation of issues in the development, deployment and use of agents as the basis for a human-computer interaction paradigm.