University of Wollongong
Browse

The sensitivity of adaptive systems to users' context

Download (16.44 MB)
thesis
posted on 2024-11-11, 13:49 authored by Wannapa Suratmethakul
The belief that computer systems should be more sensitive to users, and their context, and hence reduce the many frustrations of using these systems, is a matter of concern. The Lumiere Project, which was established at Microsoft research to address this concern, provides the background to the current research. The Lumiere Project developed User Models based on Bayesian Reasoning. The researchers of the Lumiere Project claimed that Bayesian User Models are able to infer users' needs and goals by capturing users' activities and users' queries. A system based on these models, therefore, should provide automated services at the right time when users are frustrated and need assistance from the system. Furthermore, users should be able to use common words in their queries to ask for help from the system. The current research is an exploratory study that aims to explore the effectiveness of computer systems that are sensitive to users' context. This research uses an interpretive qualitative approach to analyse data collected using laboratory experiments. These involved Usability Testing, questionnaires, and interviews. The system, chosen to be tested, is the Help incorporated in the popular Microsoft Office 2000 products. This Help System implements the User Model developed in the Lumiere Project. The results indicate that adaptive systems have limited ability to support users' goals and needs. Users will frequently ignore assistance offered by the system and can rarely find useful assistance when they seek it. Furthermore, users have difficulty using common words in help queries because the system misunderstands the words being used in the query. The research also found that some characteristics of users appear to have an influence on the effectiveness of adaptive systems. Expert users, in particular, prefer to explore menus and toolbars on their own rather than ask for assistance from the system.

History

Year

2002

Thesis type

  • Masters thesis

Faculty/School

School of Economics and Information Systems

Language

English

Disclaimer

Unless otherwise indicated, the views expressed in this thesis are those of the author and do not necessarily represent the views of the University of Wollongong.

Usage metrics

    Categories

    Keywords

    Exports

    RefWorks
    BibTeX
    Ref. manager
    Endnote
    DataCite
    NLM
    DC