CEC Theses and Dissertations


Human-Agent Interaction and Web-based Systems: A Study of User Performance and Software Agent Learning

Date of Award


Document Type


Degree Name

Doctor of Philosophy (PhD)


Graduate School of Computer and Information Sciences


Maxine S. Cohen

Committee Member

Sumitra Mukherjee

Committee Member

Timothy Ellis


The last decade has seen the Internet as an enabler to assist the user on query searching and information needs. Searching for information on the Internet differs in significant ways. To adequately support retrieval tasks, search engines require an array of emerging technologies. As supporting tools, they are of great value to the user by providing relevant results specific to a given query. This study examined Internet agents from the perspective of fixed learning and evolutionary learning with selected search engines.

Internet agents refer to the intelligent software residing in search engines to process and access information on behalf of the user. The goal of this research was to measure whether user performance varies as a result of retrieving information across selected agent types. The literature in software agents and intelligent interfaces emphasize that two fields of study guide their general development, Human Computer Interaction (HCI) and Artificial Intelligence (AI). This study applied seminal models in Information Systems as the theoretical base. Norman's seven-stage model of interaction, Simon's satisficing theory, and Goodhue's Task Technology Fit model are introduced to help explain the user interaction with Internet agents.

Two field experiments were performed to empirically test four hypotheses and answer one research question. The study was conducted at the California State University Los Angeles, Computer Information Systems department, using a sample of 60 students. User performance and personalization features were evaluated with four search engines. Using primary data, the study indicates that Internet agent types were positively associated with finding relevant results for the user. Conversely, self-evaluations on user performance were not significantly different between agent types. Research limitations are discussed as well as the contributions to the field and recommendations for future studies.

This document is currently not available here.

  Link to NovaCat