Skip to main content Skip to navigation

Towards Equity and Algorithmic Fairness in Student Grade Prediction

Project Overview

The document explores the transformative role of generative AI in education, highlighting its potential to enhance learning outcomes while addressing critical issues of equity and fairness. It focuses on the application of AI in grade prediction within higher education, emphasizing methodologies designed to mitigate racial disparities in algorithm performance. Techniques such as adversarial learning and data balancing are discussed as effective strategies for ensuring that AI systems produce equitable results. The findings underscore the necessity for educational institutions to adopt AI technologies that prioritize fairness, thereby preventing the reinforcement of existing biases. Ultimately, the document advocates for the responsible and equitable deployment of generative AI in educational settings to ensure that all students benefit from advancements in technology without facing discrimination or inequity.

Key Applications

Grade prediction models using AI

Context: Higher education institutions aiming to improve student outcomes and support historically underserved groups.

Implementation: Developed methodologies for grade prediction that include label balancing and adversarial learning to minimize bias.

Outcomes: Improved equity in grade prediction accuracy among diverse racial groups and enhanced overall predictive performance.

Challenges: Need to balance the trade-off between overall accuracy and fairness; potential for perpetuating biases if not implemented carefully.

Implementation Barriers

Bias and fairness barrier

Machine learning models may replicate and amplify biases present in historical data, leading to challenges in achieving both high accuracy and fairness in predictive models.

Proposed Solutions: Implement data processing techniques such as adversarial learning and sample weighting to mitigate algorithmic bias. Additionally, utilize strategies for group fairness evaluation and equitable sampling during model training.

Project Team

Weijie Jiang

Researcher

Zachary A. Pardos

Researcher

Contact Information

For information about the paper, please contact the authors.

Authors: Weijie Jiang, Zachary A. Pardos

Source Publication: View Original PaperLink opens in a new window

Project Contact: Dr. Jianhua Yang

LLM Model Version: gpt-4o-mini-2024-07-18

Analysis Provider: Openai

Let us know you agree to cookies