Top 36%
dans ce catalogue (796 pratiques notées)
Les scores se concentrent dans le haut de l’échelle ; la position au sein du catalogue est donc souvent plus parlante que le chiffre seul.
Evidence of impact / measured public value3/3
Lighthouse Reports and Svenska Dagbladet's 2024 analysis of a dataset of over 6,000 people flagged by the algorithm in 2017, plus IMY's formal supervision file culminating in its 18 November 2025 decision, give independently corroborated, dated evidence of both the model's effects and its fate.
Transparency, fairness & accountability1/3
Journalists needed an extensive investigative effort to obtain and test the algorithm's actual scoring logic, and it took a formal GDPR supervision case by IMY, opened after media reporting, to get Försäkringskassan to confirm in writing what the system had been doing.
Transferability / demonstrated replication1/3
Similar risk-scoring approaches are used by other European welfare agencies, but the documented bias in Sweden's own model means positive transferability has not been demonstrated.
Scalability beyond pilot2/3
The model screened applicants nationwide from 2013 until its discontinuation in 2025, over a decade of full-scale production use before being retired.
Governance, capability & sustainability3/3
An independent regulator (IMY) opened a formal case, Försäkringskassan discontinued the model and confirmed in writing it would not be reactivated, and IMY closed the case on 18 November 2025 — the clearest independent-oversight outcome of the three cases in this batch, since the practice actually stopped rather than merely being monitored.
Evidoria. L'Algorithme de Risque de Fraude de Försäkringskassan — Le Modèle Social Suédois Vieux d'une Décennie Face à Ses Propres Preuves de Biais. ID persistant: b63fc083-dfa4-48d3-8ec1-046e91b4cf59.