Part of BrusilovskysReviews.
In 1996, Brusilovsky identified 4 user characteristics to which an AdaptiveHypertext system could adapt:
However, Brusilovsky notes in the 2001 review that "a number of adaptive Web based systems are able to adapt to something [other] than user characteristics". The 2001 review identifies a further 3 sources of adaptation:
The user's knowledge of the subject represented in the HyperSpace. Many AdaptiveHypertext systems rely on user knowledge as a source of adaptation. However, a user's knowledge is variable, so system has to recognise changes and update the user model accordingly.
User knowledge is usually represented by an overlay model or a stereotype user model:
The stereotype model is simpler and easier to initialise and maintain, but less powerful than the overlay model. Some systems use the advantages of both models : the stereotype model is used to classify new users and set initial values for an overlay model which is then gradually tailored to the user as they interact with the system.
The user's goal or task as related to the context of their work in the HyperSpace. Systems often provide a set of possible user goals which they can recognise - the most appropriate goal is included in the user model.
Most changeable user feature - usually changes from session to session, often changes several times within a session.
Alternatively, the user model may hold a value for each goal supported by the system, representing the probability that a particular goal is the current goal of the user.
Background and Experience
User's background - all the information related to previous experiences outside the subject of the system, but which is relevant enough to be considered (e.g. profession, experience of work in related areas, user's point of view and perspective).
User's experience - familiarity with the structure of the hyperspace, ease of navigation within it. Not the same as users knowledge of the subject - e.g. a user quite familiar with the subject may not be familiar at all with the HyperSpace structure.
Usually modelled using stereotype model (e.g. experience stereotype, background sterotype for profession, prospect, native language).
User may prefer some nodes and links over others, and some parts of a page over others. User preferences are not deduced by the system - the user has to inform the system directly or indirectly by providing feedback.
Adaptive hypermedia systems then generalise the user's preferences and apply them for adaptation in new contexts.
Preferences are modelled numerically, whereas other parts of the model are usually represented symbolically. This allows several user preference models to be combined to form a group user model which accumalate preferences of a specific group of users. A group model can make a good starting model for a new member of the group. Group models also important in collaboration (collaborators need to see the same adapted views).
Not used in systems covered by the 1996 review, but a number of more recent Web based systems attempt to model user's long term interests in order to improve information filtering and recommendations.
Stable (long-term) user features which define a user as an individual (e.g. personality factors, cognitive factors, learning styles). Importance of user traits acknowledged, but currently little agreement on which aspects can and should be used to drive adaptation.
Adaptation to the user's enviroment brought about by Web-based systems since user can reside anywhere and be using different equipment.
Systems can adapt to the user's access platform (hardware, software, network connection) to select the type of material and media (e.g. picture vs. movie) to present the content. More advanced systems provide considerably different interfaces to different user platforms (e.g. taking advantage of user location, direction of sight and movements in handheld platforms).
-- Last edited December 4, 2002
Sponsored by Eastgate