Plataforma Tecnológica de Intervención Social — Salta's Predictive-Analytics Algorithm for Teen-Pregnancy Risk (a Cautionary Case)

In 2018, the government of Salta Province in northern Argentina, through its Ministerio de Primera Infancia, partnered with Microsoft and the anti-abortion NGO Fundación CONIN to build a predictive-analytics system intended to identify, years in advance, which named adolescent girls were most likely to become pregnant or drop out of school.

The stated aim was to flag high-risk households and individuals for early intervention using demographic scoring.

The model scored households using demographic variables including age, ethnicity, disability status, nationality, and whether the home had hot running water. Provincial 'territorial agents' then visited flagged families, photographing residents and recording GPS coordinates, feeding results back into the system.

Then-governor Juan Manuel Urtubey publicly claimed the model was '86% accurate,' but no methodology, validation data, or outcome evaluation was ever published or independently verified.

The programme drew condemnation from digital-rights researchers and academics for profiling minors by ethnicity and poverty markers without consent, transparency, or oversight, and for its ties to an anti-abortion advocacy group during Argentina's national abortion-law debate. It was quietly abandoned without a public accounting of results, and is now catalogued as a formal AI incident and analysed in peer-reviewed literature as an example of unaudited predictive profiling in social policy — illustrating why claimed accuracy figures from vendors or officials cannot substitute for independent evaluation.