Definition
Human in the loop
Also called meaningful human oversight
A requirement that a person makes or can meaningfully change an automated decision, rather than reviewing it after the fact or approving it as a formality.
Last reviewed
In practice
The word doing the work is "meaningful". A reviewer who sees a queue of pre-filtered candidates and approves the ranking has not exercised oversight, they have provided a signature. The EU AI Act is explicit that oversight must be capable of changing the outcome.
The practical test is whether the human can see and act on the cases the system would have excluded. If below-threshold candidates never reach a screen, no amount of human presence further down the process constitutes oversight of the exclusion.
This is also why auto-rejection is the specific practice regulators moved against first. It is the point at which a decision becomes unreviewable, because nobody ever learns it was made.
Related
Automated employment decision tool
A computational process that substantially assists or replaces discretionary hiring decisions, typically by scoring, ranking or filtering candidates.
Composite score
A single score combining several weighted assessment components, used to rank candidates on a common scale.
Bias audit
An assessment of whether a hiring tool produces different selection rates across protected groups, performed by a party independent of the tool’s vendor.
Read further
Longer pieces on this and the arguments around it.
What "meaningful human oversight" actually requires
A reviewer who rubber-stamps 200 rankings is not oversight. What regulators mean by meaningful human involvement, and how to build a process that holds up.
ReadThe EU AI Act and hiring: what recruitment teams are responsible for
AI used to recruit or evaluate candidates is high-risk under the EU AI Act. What that means for employers, what providers owe you, and what to ask vendors.
Read