Conference,
A Review of Clustering Models in Educational Data Science Toward Fairness-Aware Learning
(2023)
DOI: 10.1007/978-981-99-0026-8_2
Abstract
Ensuring fair access to quality education is essential for every education system to fully realize every student's potential. Nowadays, machine learning (ML) is transforming education by enabling educators to develop personalized learning strategies for the students, providing important information on student progression and early identification of potential points of struggle, developing more efficient grading systems, etc. The role of the Educational Data Science (EDS) domain in educational activities for both teachers and learners is becoming therefore increasingly important. However, ML-driven decision-making can be biased, resulting in underperforming ML models and/or ML models that discriminate against individuals or groups of students based on protected attributes like gender or race. Mitigating bias and discrimination in ML is of paramount importance. In this work, we focus on one of the most effective ML tasks, clustering, which is widely used in EDS as an exploratory tool to understand student characteristics and behavior but also as a stand-alone tool for, e.g., group assignments. Traditionally, clustering algorithms focus on finding groups or clusters of similar students and ignore aspects of fairness and discrimination. However, both cluster quality and fairness of the resulting clusters are needed. This chapter provides a comprehensive review of different clustering models in EDS, with greater emphasis on fair clustering models. Among the fair clustering models, we mainly focus on models that have been proposed and/or applied in educational activities to ensure their usefulness and applicability for fairness-aware EDS.
