In collaborative learning scenarios, the use of computers and communication networks to facilitate collaboration is becoming popular, the Computer-Supported Collaborative Learning (CSCL). In face-to-face CSCL scenarios, participants are grouped for learning activities. The location information of the participants, very useful for computational support, may involve additional configuration work for students or teachers themselves. Mobile devices that facilitate this cooperation have evolved into ubiquitous computing. Sensing devices can capture context information that allows automate the detection of the location of users and objects involved in the learning scenario and use this contextual information to improve the support offered by computers. This thesis deals in detail the problem of computational support to the detection and management of learning groups in faceto-face CSCL environments. Our research has focused on the proposal of a system that automates the management of groups in these scenarios collecting and processing contextual information from sensors. To do this we have proposed a context model for this use and we have identified what information is most relevant to this domain application. This process of modeling and identification have been theoretical ---from conceptual frameworks that have allowed us to define a model of context--- and experimental ---from assessing the quality, reliability and sensitivity of contextual information in realistic environments---. Then we have verified how this contextual information fits the contextual model. Contextual information can pass through several stages before being used. First, the contextual information collected by the sensors could be conditioned and filtered to treat quality and uncertainty. Then it is supplied to an intelligent system that learns behavior patterns of groups and students. This intelligent system requires two different operating processes: training and estimation. We have proposed specific training and assessment processes for prediction and management of groups. The output also could be conditioned as it was done with the input. Finally, we used traces of contextual information in real scenarios ---real students doing group learning activities--- to validate the system. In this validation we have taken into account both the effectiveness and their impact on the activity of students and groups. From this impact assessment we have identified patterns in the contextual information and in the behavior that have allowed us to design a system of quality, error and uncertainty management in the group estimation and a filtering system and interpolation of contextual information ambiguous, missing or erroneous, and filtering and interpolation system of ambiguous, missing or erroneous contextual information. Our thesis is that to provide computational support to the detection and management of learning groups in face-to-face CSCL environments we need three basic functionalities: 1) the collection and filtering of changes in contextual information in realtime for each student and adjust them in the context model, 2) the transformation of this contextual information and thier historical to group membership by an intelligent algorithm and 3) the quality management of group estimations to minimize the impact on activity of students because of the uncertainty of these estimations.
© 2001-2024 Fundación Dialnet · Todos los derechos reservados