CAF controls: associations demand from the government “the end of scoring algorithms”

“Put back into humanity” in the face of the excesses of dematerialization

CAF controls: associations demand from the government “the end of scoring algorithms”

“Put back into humanity” in the face of the excesses of dematerialization. Around thirty associations defending public freedoms and supporting the most disadvantaged are calling on the Prime Minister, Gabriel Attal, on the algorithm used by family allowance funds (CAF) to target their controls, in an open letter dated Monday, February 5.

This letter, signed in particular by the Changer de cap collective and the association for the defense of digital freedoms La Quadrature du Net, returns to the “discriminatory” practices generated by the algorithm for targeting CAF beneficiaries. This system, the workings of which Le Monde revealed in December, has the effect of concentrating controls on certain vulnerable groups such as single parents, some of the beneficiaries of the disabled adult allowance or low-income households.

The associations accuse this system of contributing to “institutional mistreatment” of CAFs, with “multiple material and psychological consequences”.

They therefore call on the government to abandon the use of such rating algorithms, in the CAFs but also in the rest of public organizations such as France Travail (which replaces Pôle emploi). They also demand stricter control of administrations' IT tools, as well as greater transparency about them.

Persistent opacity of CAFs

The various journalistic investigations and questioning of associations have not shaken the confidence of the National Family Allowance Fund (CNAF) in its system. “We have nothing to be ashamed of our action,” said the director of the organization, Nicolas Grivel, in an internal message sent in December, assuring that he was “neither afraid nor ashamed of the debate.”

Hearing in the Senate on January 25, the leader, however, refused to go into the details of how the algorithm works. “We are not targeting single-parent families, not at all,” he assured, explaining that vulnerable groups are more likely to be checked because they are more numerous among recipients. A line of defense which obscures the fact that the algorithm directly targets these audiences, and induces controls in greater proportions than the others.

The director of the CNAF also did not answer the question from the socialist senator of Seine-Saint-Denis Adel Ziane, who sought to know whether geographical criteria were used to target the controls. The department of Seine-Saint-Denis contacted the Defender of Rights in December, worrying about a possible “break in territorial equality”.

The CNAF has long been reluctant to shed light on the design, content and effects of its system of targeting controls. In December, Le Monde appealed to the administrative court to obtain the communication of documents relating to this “risk score”. But the public body maintains its refusal in a defense statement, signed by the law firm Veil Jourde, on January 17.

The advisors maintain that “the CNAF cannot communicate documents that it does not have”. The organization would not have established specifications before designing this algorithm, nor would it have evaluated its effects or sought to test possible biases, after more than ten years of use and hundreds of thousands of checks triggered.