Implicit Adaptation of User Preferences in Pervasive Systems

User preferences have an essential role to play in decision making in pervasive systems. However, building up and maintaining a set of user preferences for an individual user is a nontrivial exercise. Relying on the user to input preferences has been found not to work and the use of different forms of machine learning are being investigated. This paper is concerned with the problem of updating a set of preferences when a new aspect of an existing preference is discovered. A basic algorithm (with variants) is given for handling this situation. This has been developed for the Daidalos and Persist pervasive systems. Some research issues are also discussed.