New AI Framework SC2R Generates Feasible Intervention Plans for At-Risk Students

Learning analytics models flag students at risk but rarely suggest feasible interventions. A new arXiv preprint (2608.17618) proposes SC2R, a semantics-constrained counterfactual recourse framework that generates actionable, constraint-compliant intervention plans. SC2R integrates a calibrated predictive model with integer programming over discrete action variables, and uses a lightweight RDF vocabulary to represent intervention plans. SHACL validation checks timing, budget, immutability, and availability constraints. The framework was evaluated offline on the Open University's OULAD dataset, using snapshots built around each assessment at two decision horizons. Results indicate strong predictive performance, scalable generation of compact intervention plans, and that semantic validation catches infeasible plans missed by lighter optimization methods.