An Architectural System for Automatic Pedagogical Interventions in Massive Online Learning Environments

Massive Open Online Courses (MOOCs) are a teaching method that uses Virtual Learning Environments to reach a vast number of students, thus, facilitating access to education by making costs more appealing because of scale economics. Consequently, Tutors’ and teachers’ interaction is crucial for the successful development of a MOOC. However, due to the size and diversity of the student body in MOOCs, instructors and tutors need help to keep an eye on them carefully and intervene as needed. This work aims to set and validate an architecture for pedagogical interventions in online learning based on how a student feels, using the automatically detected subjective attributes obtained through interactions in the learning management systems. The architecture is based on three layers: (i) the Application layer for managing interaction with the Virtual Learning Environment; (ii) the Knowledge layer for the automatic textual classification, the attributes identification, knowledge representation through ontology and selection of pedagogical intervention actions; and (iii) the Intervention layer carries out pedagogical interventions through an autonomous conversational agent. The proposed architecture can identify the necessary pedagogical intervention, and the conversational agent can make decisions and adopt an approach more suited to the student’s needs. The proposed architecture was evaluated using the Stanford MOOC dataset, comprised of 11,042 participants who posted 29,604 messages from eleven courses. The preliminary evaluation results conclude that our approach is able to significantly support the tutor in MOOC environments as 65% of the student posts were automatically managed by the system while only the 35% left needed tutor attention.