Toward Fairly Predicting Students' Math Learning Outcomes in an Online Platform (Poster 29)

• In Event: AERA Poster Session 10
  In Poster Session: Advanced Technologies for Learning SIG Poster Session: Data Science and Artificial Intelligence in Education (San Diego)

Abstract

Educational researchers have extensively adopted artificial intelligence (AI) techniques to predict students’ learning outcomes automatically in online learning contexts. However, limited attention has been paid to the fairness of prediction with AI in educational settings that could enlarge inequality in education. This study aims to fill this gap by proposing a fair logistic regression algorithm. Specifically, we developed the fair logistic regression model and compared the fairness-aware model with fairness-unaware AI models. The results showed that the fair logistic regression algorithm could achieve comparable predictive performance while producing notably higher fairness. The implications of this study suggest that the educational community can adopt a methodological shift to achieve both accurate and fair AI to support learning and reduce bias.

Authors

• Wanli Xing, University of Florida

• Chenglu Li, University of Utah

• Walter L. Leite, University of Florida