Adapting stereotypes to handle dynamic user profiles in a pervasive system

In developing ubiquitous or pervasive systems it is essential that the complexity of the underlying system is hidden from the user. To achieve this, the system needs to take many decisions on behalf of the user. This an only be done if the system knows what the user would prefer, i.e. it maintains a set of user preferences for each user. This is a laborious task for the user to perform manually and research is focussing on the use of machine learning to assist the user in creating and maintaining an acceptable set of preferences. This paper describes how stereotypes can be adapted for use in pervasive systems to help build up user preferences while maintaining user privacy through the use of virtual identities, and how these can be modified to match the changing preferences of the group of users who select this stereotype. The paper also introduces the notion of group identities and shows how the same approach can be used to handle these in the Daidalos pervasive system.