AI IN EDUCATION, ASSESSMENT, AND INSTITUTIONAL IMPROVEMENT:

Abstract

As higher education institutions globally deploy digital infrastructures, the convergence of learning analytics, predictive modeling, and automated psychometrics is reshaping academic evaluation and institutional governance. This study examines machine learning architectures designed to track student engagement, forecast attrition, and automate assessments, evaluating their ethical alignment with United Nations Sustainable Development Goal 4 (Quality Education) across diverse geopolitical regions—including Canada, the European Union, the Western Balkans, the MENA region, and the Global South. While these technologies provide scalable instructional diagnostics, they introduce acute socio-technical vulnerabilities. Left unmanaged, automated grading engines often suffer from construct drift—evaluating dominant linguistic styles over authentic mastery—thereby marginalizing non-native student populations and regional vernaculars across the Global South and Euro-Balkan corridors. Concurrently, predictive analytics can cause structural tracking, trapping vulnerable demographics in self-fulfilling negative educational loops based on historical, localized training data.