Competences have grown in popularity in the western educational world [1, 2, 3], and so the interest on developing computational models for competences that can be used to support a variety of educational processes, from creating digital catalogues of competences to course design to monitoring competence development by students. Although meaning varies among organisations, in this paper we will assume a definition of competence along the line of'the capability of someone to act effectively in some kind of situations, which demands the mobilization of a variety of internal and external resources' which broadly integrates aspects of external performance and internal composition of competences that emerge in the literature. Research in this area is important because little information is available regarding what competences the students have developed along their studies, and to what extend, beyond the stated learning objectives of the educational programmes they are subscribed in, and the titles of the courses they have taken and passed. Furthermore, information regarding the development of competences do not accumulate, neither at school nor later in life. For example, transversal competences are develop along many courses on specific contexts (e.g.

Many real-valued connectionist approaches to learning control realize exploration by randomness inaction selection. This might be disadvantageous when costs are assigned to "negative experiences" . The basic idea presented in this paper is to make an agent explore unknown regions in a more directed manner. This is achieved by a so-called competence map, which is trained to predict the controller's accuracy, and is used for guiding exploration. Based on this, a bistable system enables smoothly switching attention between two behaviors - exploration and exploitation - depending on expected costsand knowledge gain. The appropriateness of this method is demonstrated by a simple robot navigation task.

Many real-valued connectionist approaches to learning control realize exploration by randomness in action selection. This might be disadvantageous when costs are assigned to "negative experiences". The basic idea presented in this paper is to make an agent explore unknown regions in a more directed manner. This is achieved by a so-called competence map, which is trained to predict the controller's accuracy, and is used for guiding exploration. Based on this, a bistable system enables smoothly switching attention between two behaviors - exploration and exploitation - depending on expected costs and knowledge gain. The appropriateness of this method is demonstrated by a simple robot navigation task.

Many real-valued connectionist approaches to learning control realize exploration by randomness in action selection. This might be disadvantageous when costs are assigned to "negative experiences". The basic idea presented in this paper is to make an agent explore unknown regions in a more directed manner. This is achieved by a so-called competence map, which is trained to predict the controller's accuracy, and is used for guiding exploration. Based on this, a bistable system enables smoothly switching attention between two behaviors - exploration and exploitation - depending on expected costs and knowledge gain. The appropriateness of this method is demonstrated by a simple robot navigation task.