Bridging the Gap: Evaluating AI- and Machine Learning- Driven Intervention Programs to Reduce Academic Disparities

Abstract

Academic disparities — persistent gaps in educational outcomes driven by socioeconomic inequality, geographic marginalization, first-generation status, and institutional under-resourcing — continue to undermine the equity promise of higher education globally. This paper presents a comprehensive evaluation of five categories of AI- and Machine Learning-driven intervention programs deployed across 34,800 students at five higher education institutions spanning India, the Gulf Cooperation Council (GCC), and Sub-Saharan Africa over three academic years (AY 2022–25). The interventions evaluated include AI-guided adaptive tutoring, peer mentoring networks augmented by ML-based matching algorithms, financial stress early-alert systems, AI-powered counselling chatbots, and a fully integrated AI-ML Intervention Ensemble (AIML-IE) combining all four modalities. Performance was assessed across six dimensions: academic failure rate reduction, GPA disparity gap narrowing, dropout rate change, advisor intervention timeliness, student engagement uplift, and algorithmic fairness across six demographic subgroups. The fully integrated AIML-IE achieved the strongest outcomes across all six dimensions: a 49.4% reduction in academic failure rates compared to the control group, a 50.0% narrowing of the GPA disparity gap (from 0.82 to 0.41 grade points), a 38.2% reduction in dropout rates, and equitable predictive performance across gender, socioeconomic, regional, first-generation, disability, and ethnic dimensions (minimum Disparate Impact Ratio = 0.944). Five publication-quality charts and five data tables illustrate framework architecture, longitudinal outcome trajectories, model benchmarks, feature importance rankings, and fairness metrics.