Research - A Scrutable Adaptive Hypertext

• Abstract
• SASY website
• Publications
• PhD Thesis (pdf 3.2MB)

Fuelled by the popularity and uptake of the World Wide Web since the 1990s, many researchers and commercial vendors have focussed on Adaptive Hypermedia Systems as an effective mechanism for disseminating personalised information and services. Such systems store information about the user, such as their goals, interests and background, and use this to provide a personalised response to the user. This technology has been applied to a number of contexts such as educational systems, e-commerce applications, information search and retrieval systems.

As an increasing number of systems collect and store personal information about their users to provide a personalised service, legislation around the world increasingly requires that users have access to view and modify their personal data. The spirit of such legislation is that the user should be able to understand how personal information about them is used. The literature has reported benefits of allowing users to access and understand data collected about them, particularly in the context of supporting learning through reflection. Although researchers have experimented with open user models, typically the personalisation is inscrutable: the user has little or no access to the adaptation process. When the adaptation produces unexpected results, the user may be left confused with no mechanism for understanding why the system did what it did or how to correct it.

This thesis is the next step, giving users the ability to see what has been personalised and why. In the context of personalised hypermedia, this thesis describes the first research to go beyond open, or even scrutable user models; it makes the adaptivity and associated processes open to the user and controllable. The novelty of this work is that a user of an adaptive hypertext system might ask How was this page personalised to me? and is able to see just how their user model affected what they saw in the hypertext document. With an understanding of the personalisation process and the ability to control it, the user is able to steer the personalisation to suit their changing needs, and help improve the accuracy of the user model.

Developing an interface to support the scrutinisation of an adaptive hypertext is difficult. Users may not scrutinise often as it is a distraction from their main task. But when users need to scrutinise, perhaps to correct a system misconception, they need to easily find and access the tools that enable scrutinisation of the personalisation. Ideally, the tools should not require any training and users should be able to use them effectively without prior experience or if have not used them for a long time, since this is how users are likely to scrutinise in practice.

The contributions of thesis are: (1) SASY/ATML, a domain independent, reusable framework for creation and delivery of scrutable adaptive hypertext; (2) a toolkit of graphical tools that allow the user to scrutinise, or inspect and understand what personalisation occurred and control it; (3) evaluation of the scrutinisation tools and (4) a set of guidelines for providing support for the scrutinisation of an adaptive hypertext through the exploration of several forms of scrutinisation tools.

Keywords: Open user modelling, adaptive hypertext, scrutinisation.

The following is a list of technical reports, conference and workshop papers and posters, that describe work that is part of this research and helped to shape, and provide an avenue for peer feedback of, the research submitted in this thesis.