FairSYN-Edu a Diffusion-Based Model for Fair and Private Educational Data Synthesis
Saved in:
| Title: | FairSYN-Edu a Diffusion-Based Model for Fair and Private Educational Data Synthesis |
|---|---|
| Language: | English |
| Authors: | Kadir Kesgin (ORCID |
| Source: | Discover Education. 2025 4. |
| Availability: | Springer. Available from: Springer Nature. One New York Plaza, Suite 4600, New York, NY 10004. Tel: 800-777-4643; Tel: 212-460-1500; Fax: 212-460-1700; e-mail: customerservice@springernature.com; Web site: https://link.springer.com/ |
| Peer Reviewed: | Y |
| Page Count: | 18 |
| Publication Date: | 2025 |
| Document Type: | Journal Articles Reports - Research |
| Descriptors: | Synthesis, Data, Data Science, Data Use, Artificial Intelligence, Ethics, Privacy |
| DOI: | 10.1007/s44217-025-00743-9 |
| ISSN: | 2731-5525 |
| Abstract: | The increasing demand for privacy-preserving, ethically aligned synthetic data generation in education has highlighted the limitations of existing tabular data generators. Traditional approaches often sacrifice fairness or privacy in pursuit of predictive accuracy, rendering them unsuitable for high-stakes academic settings. In this paper, we propose FairSYN-Edu, a novel diffusion-based synthetic data generation framework designed for educational data. By integrating adversarial debiasing and differentially private training into the generative process, FairSYN-Edu jointly optimizes utility, fairness, and privacy. We evaluate our approach on three real-world educational datasets spanning MOOC, K-12 tutoring, and LMS environments. Experimental results demonstrate that FairSYN-Edu achieves significantly lower demographic disparities, maintains competitive predictive performance (RMSE = 0.402), and provides moderate resistance to membership inference attacks (AUC = 0.705). Ablation studies, error gap analysis, and SHAP-based interpretability evaluations confirm the robustness and ethical reliability of our method. We release the complete implementation, synthetic benchmark suite, and documentation to promote reproducibility and responsible AI practices in education. |
| Abstractor: | As Provided |
| Entry Date: | 2025 |
| Accession Number: | EJ1482626 |
| Database: | ERIC |
| Abstract: | The increasing demand for privacy-preserving, ethically aligned synthetic data generation in education has highlighted the limitations of existing tabular data generators. Traditional approaches often sacrifice fairness or privacy in pursuit of predictive accuracy, rendering them unsuitable for high-stakes academic settings. In this paper, we propose FairSYN-Edu, a novel diffusion-based synthetic data generation framework designed for educational data. By integrating adversarial debiasing and differentially private training into the generative process, FairSYN-Edu jointly optimizes utility, fairness, and privacy. We evaluate our approach on three real-world educational datasets spanning MOOC, K-12 tutoring, and LMS environments. Experimental results demonstrate that FairSYN-Edu achieves significantly lower demographic disparities, maintains competitive predictive performance (RMSE = 0.402), and provides moderate resistance to membership inference attacks (AUC = 0.705). Ablation studies, error gap analysis, and SHAP-based interpretability evaluations confirm the robustness and ethical reliability of our method. We release the complete implementation, synthetic benchmark suite, and documentation to promote reproducibility and responsible AI practices in education. |
|---|---|
| ISSN: | 2731-5525 |
| DOI: | 10.1007/s44217-025-00743-9 |
