Accession Number : ADA320728


Title :   Utilizing Bayesian Techniques for User Interface Intelligence


Descriptive Note : Master's thesis


Corporate Author : AIR FORCE INST OF TECH WRIGHT-PATTERSON AFB OH


Personal Author(s) : Harrington, Robert A.


Full Text : http://www.dtic.mil/get-tr-doc/pdf?AD=ADA320728


Report Date : DEC 1996


Pagination or Media Count : 96


Abstract : The purpose of this research is to study the injection of an intelligent agent into modern user interface technology. This agent is intended to manage the complex interactions between the software system and the user, thus making the complexities transparent to the user. The background study will show that while interesting and promising research exists in the domain of intelligent interface agents, very little research has been published that indicates true success in representing the uncertainty involved in predicting user intent. The interface agent architecture presented in this thesis will offer one solution for solving the problem using a newly developed Bayesian-based agent called the Intelligent Interface Agent (IIA). The proof of concept of this architecture has been implemented in an actual expert system, and this thesis presents the results of the implementation. The conclusions of this thesis will show the viability of this new agent architecture, as well as promising future research in examination of cognitive models, development of an intelligent interface agent interaction language, expansion of meta-level interface learning, and refinement of the PESKI user interface.


Descriptors :   *NEURAL NETS , *EXPERT SYSTEMS , *BAYES THEOREM , LEARNING MACHINES , THESES , MAN COMPUTER INTERFACE , COMPUTER COMMUNICATIONS , SYSTEMS ENGINEERING , SOFTWARE ENGINEERING , COMPUTER ARCHITECTURE


Subject Categories : CYBERNETICS


Distribution Statement : APPROVED FOR PUBLIC RELEASE