Game theoretical, statistical and behavioural aspects of peer evaluation

• Main Author: Tan, Catherine Ser Ping
• Other Authors: Fedor Duzhin
• Format: Final Year Project
• Language: English
• Published: Nanyang Technological University 2021
• Subjects: Science::Mathematics::Statistics
• Online Access: https://hdl.handle.net/10356/148501
• Institution: Nanyang Technological University

The ”free rider” problem has long been a problem among a team of students working on a project. The most common solution to the free rider problem is peer evaluation. Although there are many systems for peer evaluation, they are prone to grade inflation or inaccurate, as they do not assign the correct grade to each student, and as a result the system is unfair. Another common concern with the current methods of peer evaluation is that students often do not have the necessary skills to evaluate the work of their peers objectively. In this paper, we introduce a new mechanism based on game theory for peer evaluation that is accurate and that does not rely on self-evaluation, i.e., if all students are completely truthful in their evaluations, then the output of our mechanism is the objective truth. Moreover, our mechanism takes into account the instructor’s judgment based on the credibility of students’ written reviews. We also construct a machine learning algorithm to identify contributions, strengths, and suggestions for improvement from a data set of peer reviews from students from the MH3110 course. This algorithm could be of convenience to instructors whose course has a large number of students, so that they do not need to take time to read the individual the peer reviews.