Date of Award

9-1998

Document Type

Dissertation

Degree Name

Doctor of Philosophy (PhD)

Department

Department of Electrical and Computer Engineering

First Advisor

Eugene Santos, Jr., PhD

Abstract

The complexity of current software applications is overwhelming users. The need exists for intelligent interface agents to address the problem of increasing taskload that is overwhelming the human user. Interface agents could help alleviate user taskload by extracting and analyzing relevant information, and providing information abstractions of that information, and providing timely, beneficial assistance to users. These agents could communicate with the user through the existing user interface and also adapt to user needs and behaviors. Central to providing assistance to a user is the issue of correctly determining the user's intent. This dissertation presents an effective, efficient, and extensible decision-theoretic architecture for user intent ascription. The multi-agent architecture, the Core Interface Agent architecture, provides a dynamic, uncertainty-based knowledge representation for modeling the inherent ambiguity in ascribing user intent. The knowledge representation, a Bayesian network, provides an intuitive, mathematically sound way of determining the likelihood a user is pursuing a goal. This likelihood, combined with the utility of offering assistance to the user, provides a decision-theoretic approach to offering assistance to the user. The architecture maintains an accurate user model of the user's goals within a target system environment. The on-line maintenance of the user model is performed by a collection of correction adaptation agents. Because the decision-theoretic methodology is domain-independent, this new methodology for user intent ascription is readily extensible over new application domains. Furthermore, it also offers the ability to "bootstrap" intent understanding without the need for often lengthy and costly knowledge elicitation. Thus, as a side benefit, the process can mitigate the classic knowledge acquisition bottleneck problem.

AFIT Designator

AFIT-DS-ENG-98-12

DTIC Accession Number

ADA354267

Download

Share

COinS