Towards a Tutoring Method in CSCL Environments through Prediction of Analysis Indicators using Generalized Linear Models

The COLLECE system is a CSCL (Computer-Supported Collaborative Learning) environment designed for learning programming through a problem-solving approach. It has been used as a research prototype in educational and professional settings. This system includes an analysis module that allows for characterization of both the collaborative process and the programs built (solutions). This work proposes a method based on generalized linear models to provide support to users during their collaborative learning by (i) utilizing a set of analysis indicators, (ii) predicting the values of these indicators during the activity, and (iii) presenting this information to users so that they can redirect their activity and achieve better outcomes in terms of the process to follow and the characteristics of the product being developed. This method is exemplified through its application to a set of activities conducted with COLLECE, resulting in the finding that the method enables prediction of a majority of indicators in Java programming teaching activities of different complexities and natures before these activities conclude.