User Model Agents for Distributed Filtering

This paper describes an architecture for intelligent agents derived from user models. The work extends our previous work on user modelling as a basis for various forms of customised systems. We make use of the um toolkit for representation, management and acquisition of the user models. We hav e used um-models as the basis for customisation of a coaching system, a personalised hypertext for teaching C and a movie advisor The system described here extends the application of such user models by constructing intelligent agents from domain-dependent parts of user models and distributing these agents to publication centres where they filter newly published objects.