IMPLEMENTING A CONTEXT-SENSITIVE MOBILE LEARNING SYSTEM Anders Kofod-Petersen, Gunhild Griff Bye, John Krogstie IDI, NTNU, Trondheim, Norway anderpe@idi.ntnu.no ABSTRACT As learning moves beyond the classroom and into everyday life, the supporting technology must be able to adapt, not only to the situations in which learners wish to learn but also their particular style of learning. Modern learning para- digms, such as mobile and ambient learning embrace adaptation through context-sensitivity. Context-sensitivity allows systems to carry out their main functions, whilst still adapting their behaviour. A user’s idiosyncrasies are important contextual information that must be handled. The work presented here demonstrates how Dreyfus’ level of competence, Gardner’s multiple intelligences and Hofstede’s cultural dimensions can be used as a basis for modelling users through stereotypes, and further how a context-sensitive mobile learning systems can use this to adapt its behaviour when assist- ing learners. It shows how an implementation of such stereotypes and details initial functionality testing. KEYWORDS Mobile learning, user modelling, context. 1. INTRODUCTION Advances in ubiquitous and mobile technology have facilitated learning outside the classroom, where and when desired. Technology now allows learners to access recourses and interact with their peers and teachers while being mobile. This mobility heavily influences the development of suitable learning system. Many learners wish to continue their learning in response to ongoing situations without time being specifically set aside for it [Eraut, 2006] or combining it with leisure [Thornton et al., 2005]. The flexibility of modern learning paradigms, in particular mobile learning [Sharples et al., 2005] and ambient learning [eTEN, 2004], fosters a closer look at making the corresponding systems context-sensitive. Learning can be claimed to be situated [Greeno, 1998]. Some of the most important contextual information in mobile learning systems is the competence of the users and their style of learning. Constructing a system that is sensitive to this type of information requires a user model. The type of user model is dependent of whether the user group is highly homogeneous or heterogeneous. In the former case the easiest solution is using ca- nonical user models, whereas in the latter specific user models are likely to be preferred. With respect to the specific learning styles, mobile learning appears to be leaning towards the use of a specific user model. How- ever, specific user models can often be a time consuming, both in modelling and run time. The work presented here argues that the use of stereotype user modelling is a promising approach for constructing feasible general, yet sufficiently specific user models. We take our departure from the work by Rich [1983] on stereotype modelling, which we attempt to ground in learning and social theories from Drey- fus [1998], Gardner [1984] and Hofstede [2001]. The rest of the paper is structured as follows: First, the relevant background is described. Then, an over- view of our design and implementation is presented. This is followed be a running example from our initial testing of the system. Finally, the paper is concluded and some pointers to future work are presented.