Glossary
Algorithmic impact assessment
A structured review of an automated system's possible effects on people, done before or while it is used.
An algorithmic impact assessment asks what an automated or AI system will do, who it could affect, what could go wrong, and how risks will be reduced and monitored. Some governments require them for public-sector systems, and some laws require them for systems used in decisions about jobs, housing, credit, or services.
Related tools include algorithm registers, public lists of the automated systems an agency uses. Both give journalists and the public a starting point for asking how a system works.
Assessments are often completed by the organization deploying the system, and their depth varies widely. Publishing one shows a process was followed, not that the system is fair or accurate.
Read more: our explainer