An Environment to Model Massive Open Online Course Dynamics

Learning at every time and in every place is nowadays possible thanks to the exponential growth of the Internet and of services deployed through it. Due to its undeniable advantages, Distance Education is becoming strategic in many fields of daily life, and encompasses both educational as well as training applications. Present platforms suitable for the former include Moodle, ATutor and others. Coursera is a popular example of a MOOC-type (Massive Open Online Courses) platform that offers different courses to thousands of enrolled students. Like happens for other technological advancements, there is also a reverse of the medal. As a matter of fact, new problems arise, such as the reliable assessment of the learning status of the learner. This is a critical point especially when the assessment has an academic/legal value, and becomes dramatic when thousands of students attend a course, as is in MOOCs. In these cases, Peer Assessment, possibly mediated by a light teacher’s activity, can represent a valuable solution. The evaluation mostly involves peers, and further creates a kind of dynamics in the community of learners that evolves autonomously. Such evolution can provide further information on both individual and group learning progress. This paper proposes a first step along this line, which entails a peer assessment mechanism led by the teacher. However, the latter only enters the process by evaluating a very small portion of the students. The proposed mechanism relies on machine learning, and in particular on a modified form of K-NN. Given the set of teacher’s grades, the system is designed to converge towards an evaluation for the remaining students, that is as similar as possible to the one that the teacher would have given. The results of the presented experiment are encouraging and suggest more research on the topic