Using Graph Embedding to Monitor Communities of Learners

Abstract

How to keep track of the learning process of a community of learners is a problem whose resolution requires accurate assessment tools and appropriate teaching and learning strategies. Peer Assessment is a standard didactic strategy which requires students in a course to correct their peers’ assignments. Since the representation of a community, even a large one, of students, is based on directed graphs, it is difficult to follow its whole dynamics. In this paper, we investigate the possibility of using two machine learning techniques: Graph Embeddings, and Principal Component Analysis, to represent a students’ communities by points in a 2D space, in order to have valuable and understandable information on the dynamics of the group. For this purpose we present a case study based on three real Peer Assessment sessions. The first results are encouraging.