Abstract:

Research on automatic personalization (or "adaptation") of information sources, mostly in an educational setting, has resulted in a number of platforms that consist of a webserver extension that performs user modeling and adaptation using server-side code and storage. There are several reasons why this server-side approach is not the best, including reasons of performance (all computation to be done on a server) and privacy (all user data is stored on the server).

At the TU/e a research project was started, called WiBAF (Within-Browser Adaptation Framework), that has been adapted to form the technological basis for the Autism&Uni toolkit. WiBAF uses browser-side storage and javascript to perform the most privacy-sensitive tasks of user modeling and adaptation on the end-user's machine. Apart from login/authorization data nothing about the user is stored on the server side. For students with some degree of autism privacy is a major concern. With WiBAF they can receive personalized information without revealing their personal characteristics to the information server.

For the Autism&Uni project the adaptation has been incorporated in a popular Wordpress content-management system to enable authors to create information once and have it be automatically adapted to each individual user. Two dimensions of adaptation have been realized in WiBAF for this purpose: adaptation to the analytic vs. global learning style dimension and adaptation to the visual vs. verbal learning style. Adaptation using other dimensions can still be added. We are currently considering adding also the active vs. reflective dimension.

In the Autism&Uni project introductory information is being created for students entering the university for the first time. It includes scenarios for the different processes, administrative or physical, such as finding your way around the campus. For the analytic vs. global learning style dimension WiBAF adaptation code can either present the information in a piecemeal fashion, using a series of slides, or present the whole story for a scenario all at once. For the visual vs. verbal dimension WiBAF re-orders fragments, to either start or end with visual information such as a comic strip and/or a video. WiBAF can also be configured to leave out elements or to scale down visual information for verbal users. As an example of the combination of adaptation strategies, to present "directions" to a location on the campus we can (depending on the material available) provide the student with written directions (step by step for a verbal analytic user or a whole list at once for a verbal global user), a combination of instructions with images (still step by step but for a visual analytic user), or by means of a map (for a visual global user).

When re-ordering information, WiBAF can also take prerequisites into account. An author can indicate that certain steps or scenarios must precede others and the adaptation will then preserve that order.

The adaptation platform makes it possible to share the user model data with other applications, for instance with course material offered through Wordpress. Whether the student wishes to share personal data with these other applications is kept under the user's control so as to satisfy also the more privacy-concerned students.

Keywords:

Autism, open source, disability, adaptation, learning styles, personalisation.