AI Governance Engineer
GitHub

Human oversight, designed

Actions classified by consequence, a designed checkpoint only where the stakes justify it, and oversight that leaves evidence which is itself monitored.

By Updated First published

Human oversight, designed Between reviewing every action and rubber-stamping all of them, actions are classified by consequence, a designed checkpoint with context sits only where the stakes justify it, and the oversight leaves evidence that is itself monitored. Art. 14: understand, decide against, stop Undifferentiated, it fails both ways Review every action destroys the value Nominal review a rubber stamp Proposed action Consequence? low Proceeds high Designed checkpoint Context to disagree Layer 04 Runtime Controls & Observability Oversight evidence Approver and decision, logged Approval rate, time-to-decide and override rate, watched as a signal Oversight you do not measure is oversight you cannot claim.
Human oversight, designed Human oversight as an engineered control: classify each action by consequence, put a designed checkpoint only where the stakes justify the latency, and keep the decision and the oversight metrics as evidence. Decide which actions need a person by consequence, then watch approval rate, time-to-decide and override rate. Drawn from chapter 04.

Text alternative

Article 14 requires that a person can effectively oversee a high-risk system: understand the output, decide against it and stop the system. Undifferentiated oversight fails both ways: human review of every action destroys the value of automation, and nominal oversight of a firehose of actions is a rubber stamp. The engineered answer classifies each proposed action by consequence and places a designed checkpoint only where the stakes justify the latency, in Layer 04 Runtime Controls & Observability (the Human-in-the-loop Gate pattern); other actions proceed without one. The reviewer gets enough context to disagree, which designs against automation bias. The approver and the decision are logged as evidence, and the oversight itself (approval rate, time-to-decide, override rate) is monitored as a signal that can degrade. Oversight you do not measure is oversight you cannot claim.

Download

Every file carries the attribution band "aigovernanceengineer.com · CC BY 4.0 · v0.5.0" inside the image, and the version is in the file name, so a copy always says where it came from and which edition it shows. The SVGs keep the text live: the first follows the viewer's light or dark setting, the other two fix one theme for slides and print. The PNGs are drawn with the site's own typefaces.

Actions classified by consequence, a designed checkpoint only where the stakes justify it, and oversight that leaves evidence which is itself monitored.
The PNG, light, 1600 px wide, as it downloads (shown here as a lighter copy).

Reuse and credit

The figure is published under CC BY 4.0: you may copy, share and adapt it, commercially too, provided you give appropriate credit, link to the licence and say if you changed it. Keep the attribution band in the image. A credit line that covers title, author, source and licence:

“Human oversight, designed” by Jorge García Aibar, aigovernanceengineer.com (https://aigovernanceengineer.com/figures/human-oversight), v0.5.0. Licensed under CC BY 4.0 (https://creativecommons.org/licenses/by/4.0/).

Embed with HTML

<figure>
  <img src="https://aigovernanceengineer.com/downloads/figures/human-oversight-v0.5.0-light-1600.png" alt="Actions classified by consequence, a designed checkpoint only where the stakes justify it, and oversight that leaves evidence which is itself monitored." width="800" height="1137" loading="lazy">
  <figcaption>
    <a href="https://aigovernanceengineer.com/figures/human-oversight">Human oversight, designed</a> by Jorge García Aibar,
    aigovernanceengineer.com, v0.5.0.
    Licensed under <a href="https://creativecommons.org/licenses/by/4.0/">CC BY 4.0</a>.
  </figcaption>
</figure>

Embed with Markdown

![Actions classified by consequence, a designed checkpoint only where the stakes justify it, and oversight that leaves evidence which is itself monitored.](https://aigovernanceengineer.com/downloads/figures/human-oversight-v0.5.0-light-1600.png)

*[Human oversight, designed](https://aigovernanceengineer.com/figures/human-oversight) by Jorge García Aibar, aigovernanceengineer.com, v0.5.0. Licensed under [CC BY 4.0](https://creativecommons.org/licenses/by/4.0/).*