No consequential automated judgment without notice, reasons, accountable human appeal, and the power to repair.
Tell the affected person that an automated or AI-enabled system materially influenced the outcome. Identify the system’s role, responsible institution, decision date, and durable case reference.
A vague reference to “technology” is not notice.
State the principal data, rules, objectives, material factors, source records, and inferences that shaped the result in language a person can understand.
“The algorithm decided” is not a reason.
Provide a named or identifiable reviewer with time, independence, relevant information, and real authority to reconsider and reverse the outcome.
A chatbot loop is not an appeal.
Correct inaccurate source data, add missing context, notify downstream systems, reverse the result where justified, and restore lost access, pay, status, opportunity, reach, reputation, or standing.
A correction without downstream repair leaves the machine’s lie alive.
The language changes by sector. The public standard does not.
The demand becomes public power when people can repeat it, compare outcomes, and show which institutions corrected the record—and which refused to answer.
