France Travail: AI That Classifies Job Seekers Raises Concerns

⚡

Key Takeaways

1France Travail is testing an AI to classify job seekers into "suspect" and "non-suspect" profiles.

2La Quadrature du Net revealed a document about this tool, highlighting risks of dehumanization.

379% of the cases targeted by the AI have been reviewed, compared to 53% previously, with 64% resulting in sanctions.

💡Why it matters — The use of AI to automate checks could impact millions of job seekers, raising questions of transparency and ethics.

An AI to Sort Job Seekers

France Travail, the organization responsible for employment in France, is developing an artificial intelligence (AI) tool aimed at sorting job seekers. The stated goal is to automatically classify profiles to enhance controls while avoiding human biases. This AI could thus intensify the existing system of random checks.

La Quadrature du Net Sounds the Alarm

The organization La Quadrature du Net revealed the existence of this project by publishing an internal document from France Travail titled "Targeting Employment Search Control." This document details how the AI is used to identify profiles deemed "suspicious," meaning those that will be subjected to more rigorous checks. Currently, the tool is being tested on job seekers who have signed a mutual termination agreement, but its use could extend to all registered individuals.

The AI's Criteria for Classifying Profiles

France Travail's AI examines each file based on 26 specific criteria. Among these criteria are the activities declared by the job seeker, the visibility of their CV online, their level of education, and the length of their job search. After analysis, profiles are classified into two categories: "suspicious" and "non-suspicious." Files deemed "suspicious" become a priority for checks, which can lead to sanctions or removal from the system.

An Accelerated Control Process

Although France Travail already has control procedures in place, this tool aims to speed them up. The institution justifies the use of AI by the desire to eliminate human "preconceptions," thus making checks more "objective." In other words, the AI analyzes files impartially, without human intervention, relying solely on objective data.

Promising Results According to France Travail

According to France Travail, the tool has already demonstrated its effectiveness. Indeed, 79% of the files targeted by the AI underwent thorough examination, compared to only 53% with the previous random selection system. Furthermore, 64% of the files checked resulted in sanctions or a "re-energizing" of job seekers.

Concerns About Transparency and Ethics

Although the AI is presented as impartial, La Quadrature du Net criticizes the lack of transparency, particularly because the source code of the tool is not public. This means that France Travail can adjust the criteria for suspicion at will, without external oversight. The institution has also acknowledged a risk of bias against low-income individuals and has adjusted the algorithm accordingly.

A Potential Impact on Millions of People

This new tool could affect more than 6 million job seekers in France. The idea is to minimize human intervention in the control and job search processes, an approach already adopted by the Caisse nationale des allocations familiales (CNAF).