FairEduNet Algorithm Design and Teaching Assessment Calibration Considering Equity in Higher Education

Area of Science:

• Educational Technology
• Artificial Intelligence in Education
• Data Science

Background:

• Current teaching assessment calibration methods struggle with integrating multi-source heterogeneous data, leading to low efficiency.
• Existing methods lack adaptability across different disciplines and teaching scenarios, and exhibit fairness biases.
• These limitations hinder the core requirement of balanced teaching in intelligent education.

Purpose of the Study:

• To construct an interpretable, transferable, and scalable fairness-correction framework for teaching assessment data.
• To develop a deep modeling and dynamic correction approach for educational assessment data.
• To address the challenges of data integration efficiency, adaptability, and fairness in teaching assessments.

Main Methods:

• Proposed a fairness-oriented educational network algorithm integrating a generative adversarial network (GAN) and a gradient boosting decision tree (GBDT).
• Developed a fairness calibration mechanism using a bidirectional encoder representation model (BERT) and a graph attention network (GAT) for feature extraction and dynamic adaptability.
• Implemented a framework with three core modules: multi-source data preprocessing, dynamic fairness index verification, and explainability visualization.

Main Results:

• Achieved 98.76% accuracy in clustering evaluation features and 98.05% in fairness correction on the validation set.
• Demonstrated 0.968 cross-scenario correction consistency and a reduction in total loss from 0.1237 to 0.1018.
• Outperformed comparative models in multi-source data integration and fairness bias correction.

Conclusions:

• The proposed fairness-oriented educational network algorithm effectively addresses limitations in existing teaching assessment calibration methods.
• The model enhances multi-source data processing efficiency and fairness calibration accuracy, promoting teaching equity.
• This research offers innovative algorithmic frameworks for developers and reliable tools for educational administrators.