Policy & Safety

The right to a human: appeal rules for algorithmic decisions take hold

New requirements guarantee human review of consequential AI decisions in lending, hiring, housing and benefits — and early data shows the appeals matter.

By Sofia Lindgren, AI Policy Correspondent — BERLIN

BERLIN — The newest right taking shape in the algorithmic age is an old one: the right to be heard by a person. Rules now in force across several jurisdictions guarantee human review of consequential automated decisions — loan denials, job screening rejections, benefits cuts — and the first year of data suggests the safeguard has teeth.

Roughly one in seven appealed decisions is being overturned, early figures show, most often where automated systems misread nonstandard circumstances: gig income, caregiving gaps, informal housing.

Institutions initially feared appeal floods that mostly did not come; the deeper effect, administrators report, is upstream — systems are being redesigned to avoid generating appealable errors in the first place.

Enable JavaScript to read the full story on Neural Daily News.