Human to output
Oversight
- 0Every output reviewed in full
- 1Reviewed before effect
- 2Per-action approval
- 3Sampled or exception-based
Fidarium · The Proximity Method
Three numbers from 0 to 3 show how close each AI capability stands to human oversight, to its governing document and to decisions about people.
Human to output
Capability to governing document
Output to a decision about a person
Declared by a person on the registration card. Binding.
Observed from runs, logs, the document chain and the quality ledger.
Estimated. Expires at the next review.
A Discovered value that disagrees with the Defined value opens a finding; it never silently overwrites it.
No promotion while any value is Approximated.
Profile 3-3-0 is not permitted: far from oversight, ungoverned, and making decisions about people.
Example profile
A person checks every output before it takes effect, but no document governs this capability yet, and it informs a decision about a person. Fix the orphan first; autonomy stays low.
Find your own profile