France Travail Algorithm Flags Job Seekers as 'Suspect' Based on 26 Variables
France Travail, the French national employment agency, is utilizing an algorithm to identify job seekers for priority control, based on 26 distinct variables. This system, internally named 'Ciblage du Contrôle de la Recherche d'Emploi' (Targeting Employment Search Control), analyzes past data from 60,000 controls to categorize individuals as either 'suspect' or 'non-suspect.' Factors that can trigger scrutiny include having an online CV on the France Travail website, working in a sector with high demand for labor, or even missing an appointment two years prior. Those flagged as 'suspect' are placed on a list for prioritized monitoring. The existence and details of this algorithm were revealed by the digital rights group La Quadrature du Net in collaboration with Radio France's investigative unit. The document detailing the algorithm was presented to the France Travail AI ethics committee on December 10th.
The implementation of an algorithm by France Travail to flag job seekers as 'suspect' based on 26 variables raises significant questions about fairness and due process in public services. While leveraging data for efficiency is a common practice, the use of a decision tree trained on past controls, which categorizes individuals, could perpetuate existing biases or unfairly penalize individuals for past minor infractions. The opacity surrounding the precise weighting of these 26 variables and the potential for algorithmic errors necessitates robust oversight. Future iterations should prioritize transparency, provide clear recourse for individuals flagged by the system, and ensure that the algorithm's design does not disproportionately impact vulnerable populations, thereby maintaining public trust in employment support systems.
AI-generated to prompt reflection — not editorial opinion, not advice, not a statement of fact. How this works.