Simulating massive open on-line courses dynamics

Learning without constraints of space and time has nowadays become a reality thanks to the exponential growth of the Internet, and distance learning is one of the areas for which that growth has been the most beneficial. The Web offers several platforms for distance learning, calibrated for the public or private sector. Among these, the Massive Open Online Courses (MOOCs) are one of the most popular online learning assets. In MOOCs, a crucial feature is given by the amount of enrolled learners: thousands can attend such courses. Consequently, monitoring the learning process of individual students is a challenging task for teachers in MOOCs. Peer Assessment can be used, in these case, as a teaching strategy, to follow the dynamics of a MOOC while not checking directly on each one of the individual students. In previous work we have presented a method to support peer assessment, and grades inference, mediated by a limited grading work performed by the teacher. The method is based on a definition of student model based on a modified version of the machine learning technique called K-NN. In this paper we present a system able to simulate the dynamics of a MOOC class where peer assessment, mediated by the teacher, takes place. The system supports the creation of a (big) class, and the simulation of peer assessment sessions, where peer grading, and teacher grading take place. The system, then, allows for a statistical analysis of how the student models computed by the system, and the grades inferred for the peers' tasks, converge towards the real values for the student models and grades.