France Travail: AI to classify “suspect” jobseekers for checks

France Travail is working on an AI system intended to sort jobseekers deemed “suspect”. The tool is being introduced to target checks more effectively.

France Travail AI targets jobseeker checks

Artificial intelligence is driving a further loss of human involvement. France Travail is developing an AI tool designed to sort jobseekers. The plan is to automatically categorise profiles in order to step up checks while avoiding any human “preconceptions”. The AI would therefore reinforce the existing random-check system.

The warning was raised by La Quadrature du Net, which has revealed an internal working document from the organisation, entitled “Targeting Checks on Job Search Activity”. It describes a tool using AI to identify “suspect” profiles - in other words, those who will, or will not, be checked. It is currently being tested on registered claimants who agreed a mutually agreed termination of their employment contract, though it could eventually be rolled out to everyone.

AI will assess jobseekers’ engagement

The AI will examine every record in its database against 26 specific criteria, including declared activities, the online visibility of a CV, education level and length of service. Each profile will then be placed in one of two groups: “suspect” or “non-suspect”. “Suspect” cases will become priorities for checks and may consequently result in removal from the register.

Checks already existed at France Travail, but the purpose of this tool is to speed up the process. To defend its use, the organisation cites “the removal of preconceptions” to make checks more “objective”. In practice, the machine is intended to assess records dispassionately, without prejudice, fully removing human involvement from the process and relying solely on raw data. France Travail says that this approach has already proved effective: 79% of files selected by the tool led to a more detailed review, compared with 53% under the former random-selection system. Furthermore, 64% of checked files resulted in penalties or “re-engagement”.

Concerns over the France Travail algorithm

On paper, this is an impartial AI that analyses data to identify registered claimants considered suspect. Beyond the questionable nature of this approach, La Quadrature du Net criticises the fact that the source code has not been made public. France Travail can alter the threshold at which a file is classed as suspect as it sees fit. The organisation has also identified the risk that checks could disproportionately focus on people with low incomes, and the algorithm was amended accordingly. In total, more than 6 million people could be affected by this new tool.

The underlying aim is to remove the human factor as far as possible in order to automate processes, whether for checks or job searches. This method is already used by another organisation: the National Family Allowances Fund (CNAF).