Video of my lecture on User Modeling on Distributed Environments given at the summer school on adaptive educational system ETIAH,Grenoble, France, July 7-11, 2007.

Workshop on Ubiquitous and Decentralized User Modelling (UbiDeUM'2007), held at UM'2007 in Corfu, Greece, 26 June, 2007 (organized by Shlomo Berkovsky, Keith Cheverst, Peter Dolog, Dominik Heckmann, Tsvi Kufik, Jerome Picault, Phivos Mylonas, and Julita Vassileva.

WORKSHOP on Decentralized, Agent-Based and Social User Modelling Approaches (DASUM'2005), held with UM'05 in Edinburgh, 25 July, 2005 (Julita Vassileva and Peter Dolog, co-organizers)

It is common to think of a "user model" as the system's representation of certain user characteristics. In a distributed multi-agent based software environment, however, this traditional user model seizes to exist and is replaced by user model fragments, developed by the various software agents, populating the environment. These model fragments have been developed in a variety of specific contexts and with various goals. User models are thus fragmented, relativized, local, and often quite shallow. They are inherently inconsistent with one another and reflect not only characteristics of the users, but also certain social relationships among them. In addition, depending on who is modelling and who is being modelled in a distributed multi-agent environment, there can be agents modelling users, users modelling agents and agents modelling other agents. With the arising wealth of models and data, the user modelling problem transforms into retrieval and integration of the available user model fragments "just in time" by a particular computational agent to the breadth and depth needed for a specific purpose. Defining ontologies of purposes, together with relevant sources of user modelling data and mechanisms for retrieving and integrating user model becomes a major research goal.


The purpose-based user modeling approach in one form of decentralized user modeling aims to develop a library of purposes - that define context dependent user modelling procedures. Each purpose has input (what types of user data it needs, from which sources the data should be requested), procedure (how to combine and proces the input) and output - either adaptation decision, or new user data to be added to the user profile. This approach allows for a gradually expanding set of user modelling processes, each handling in context, for particular goal, a variety of contextualized user data from various sources.


In online communities, the problem arises how to support users to migrate seamlessly from community to community, to move their user data, to adapt to their participation and reputation accumulated in different communities. The domain is ideal for decentralized user modeling, since communities are autonomous, using their own user modeling and representation techniques. User profile fragments exist in many different communities, created and interpreted in the contexts of these communiteis. A variation of the purpose based approach can be applied through designing policies, which are equivalent to purposes - procedures that collect the user data required, process it in context and make decisions for adaptation or generate new user data. The policies are open and can be edited by users in specific roles (e.g. owner of the community). This allows for user participation in the modeling process i.e. "participative user modelling".