Predicting Priority Cases for Uruguay's Crece Contigo — An MIT-MIDES Fairness-Aware Model Still Under Evaluation

Uruguay's Ministry of Social Development (MIDES) runs Uruguay Crece Contigo, supporting pregnant women and children under four living in extreme poverty; caseworkers historically prioritised referrals manually, risking inconsistency and bias. MIT students, advised by a PhD candidate, worked with MIDES on anonymised data from over 15,000 historical referral cases, testing gradient-boosted trees, neural networks, LSTMs and ensemble methods with fairness-aware techniques (SMOTE, RUS) intended to reduce bias against under-represented groups.

As of mid-2025 the model remained a decision-support prototype under internal MIDES evaluation; no deployment date or performance figures have been published, making this an early-stage, honestly-unproven case rather than a demonstrated success story.

It is nonetheless Uruguay's first documented attempt to apply machine learning to social-programme targeting, and reflects the country's broader push -- a national AI Observatory and AGESIC oversight -- toward auditable government AI use.