Efficient Lecturer Peer-Assessment Attribution Using Graph Theory and a Novel Greedy Algorithm

This paper introduces an efficient algorithm to address the issue of lecturer peer-assessment assignment. The motivation for this solution arises from a real-world scenario where a group of lecturers receives teaching feedback through the process of peer assessment. In this context, in the case of a large group, manually keeping track of desires and constraints is hard, and therefore a computational solution is paramount. The proposed technique looks for a solution where every teacher is evaluated by a target number of peers. Moreover, affinity between peers may be encoded in the algorithm to give preference to solutions where the assignments have higher affinity. The problem is framed using a directed weighted graph, where the weights are the affinity between peers, and the proposed greedy algorithm regularizes this graph to achieve the attribution. Results are presented where the proposed approach is applied to both real and simulated data, resulting in adequate attributions within an efficient time frame.