ELI–SDG-Net: An Explainable AI Framework for Data-Driven Ethical Leadership Indexing and SDG-Based Institutional Ranking in Higher Education
Year: 2026
Type: article
Abstract: This study introduces Ethical Leadership Index-Sustainable Development Goal-Net (ELI-SDG-Net), a unified deep learning framework that derives ELI through a component-weighted autoencoder trained on governance, faculty diversity, pay equity, and student outcome variables drawn from institutional administrative data. The framework formulates a novel joint training objective, Ltotal, which co-optimizes latent representation quality and outcome prediction, and incorporates a graph-based multi-task module that jointly predicts all 17 SDG scores using institutional embeddings. An experimental evaluation was conducted on a merged dataset of 4,821 institutions drawn from the IPEDS and Times Higher Education datasets. The merged dataset was assessed against five baseline models using Root Mean Square Error (RMSE), R2, and Mean Absolute Error (MAE). The proposed framework achieved an RMSE of 0.089 and an R2 of 0.91, exceeding the next-best baseline by 0.087 RMSE points and outperforming prior survey-based SEM studies by over 43 percentage points in explained variance. The multi-task module reduced prediction error on 15 of 17 SDG goals relative to single-task baselines, and the SHAP analysis confirmed graduation rate, faculty pay equity, and administrative transparency as the primary governance drivers. ELI-SDG-Net establishes a reproducible, interpretable foundation for quantitative governance evaluation at the global institutional scale in higher education.
DOI: 10.48084/etasr.19072
Authors: Sowmya C. U., Smitha V. Shenoy
Language: English
Subfield: Education
Field: Social Sciences
Domain: Social Sciences
Open Access status: gold
Estimated APC paid: $431
OpenAlex ID: W7202275428