Towards an Extended Model of User Interface Adaptation: The Isatine Framework

In order to cover the complete process of user interface adaptation, this paper extends Dieterich's taxonomy of user interface adaptation by specializing Norman's theory of action into the I satine framework. This framework decomposes user interface adaptation into seven stages of adaptation: goals for adaptation, initiative, specification, application, transition, interpretation, and evaluation. The purpose of each stage is defined and could be ensured respectively by the user, the interactive system, a third party, or any combination of these entities. The potential collaboration between these entities suggests defining additional support operations such as negotiation, transfer, and delegation. The variation and the complexity of adaptation configurations induced by the framework invited us to introduce a multi-agent adaptation engine, whose each agent is responsible for achieving one stage at a time (preferably) or a combination of them (in practice). In this engine, the adaptation rules are explicitly encoded in a knowledge base, from which they can be retrieved on demand and executed. In particular, the application of adaptation rules is ensured by examining the definition of each adaptation rule and by interpreting them at run-time, based on a graph transformation system. The motivations for this multi-agent system are explained and the implementation of the engine is described in these terms. In order to demonstrate that this multi-agent architecture allows an easy reconfigurability of the interactive system to accom modate the various adaptations defined in the framework, a case study of a second-hand car-selling system is detailed from a simple adaptation to progressively more complex ones.