Software Agents for Personalized Training of Collaborative E-learning Teachers
Pablo Santana Mansilla · International Joint Conference on Artificial Intelligence · 2014
Introduction and Proposal. In Computer Supported Collaborative Learning (CSCL) environments it is not enough to organize students in groups because interactions that originate the collaborative creation of knowledge do not happen spontaneously [9]. Consequently, it is essential that e-tutors (teachers) promote interactions between students and coordinate discussions [9]. To ease the workload (both temporal and cognitive) that supposes for teachers the coordination of discussions, software applications that give information about critical events of the interaction process have been developed so that teachers can intervene when group conflicts take place [9]. Most e-tutors lack of experience with techniques of CSCL [6], but coordination skills of CSCL interactions belong to acquired skills category, it means this kind of skills can be developed through training processes [3]. Software applications that assist CSCL e-teachers have recurred to structured interfaces and free text in order to model interactions. Structured interfaces (such as sentence openers) simplify the process of interactions analysis, yet there are studies that demonstrate this kind of interfaces restrict interaction possibilities and cause slow communication [4]. Considering the need for mechanisms to train e-tutors and the existence of shortcomings in structured interfaces, it is proposed the creation of a multi-agent model for personalized training of e-tutors in CSCL environments where communication is based on a free text interface. In the proposed model there will be some agents who will analyze interactions between students (based on free text) in search of conflict situations whose resolution requires the intervention of e-tutors. In these cases, other agents will suggest e-tutors actions to take so as to solve group conflicts and simultaneously train those skills they have not properly manifested. Related Work. Software applications related to the proposed model, in the sense they suggest actions according to analysis of recorded interactions, are: ARGUNAUT [9], I-MINDS [8] and SAVER [5]. However, none of the named applications considered skills of e-tutors. The research line of Performance Support Systems [3] should also be taken into account to establish strategies and mechanisms commonly used to promote the acquisition of skills. On the other hand, in the context of CSCL the use of natural language processing (to analyze interactions based on free text) has been limited to tasks such as classification of conversation topics or characterization of discussion patterns [9]. Doctoral Consortium IJCAI School 2014