One structural distinction has to be made before governance can work:
When someone leaves a governed composition, did an EXIT happen — or did the composition SPLIT?
That question sounds administrative. It is not.
In AI governance, federated systems, long-lived digital entities, and human-AI continuity architectures, this distinction becomes structural.
If one participant, anchor, role, or authority-holder leaves, the system must not simply say:
“the rest continues.”
It also must not dramatize every departure as a split.
The threshold is narrower:
Does exactly one lawful successor composition remain?
If yes, the transition may be EXIT.
If multiple incompatible successor claims remain, it is SPLIT.
If evidence, standing, authority, witness lineage, obligation status, or temporal window state is unclear, the correct state is HOLD / FREEZE.
If no admissible successor remains, it may be DISSOLVE — but only after witnessed successor-space exhaustion.
That is the purpose of the A6 Composition Transition Predicate Addendum v0.1.4.
It makes explicit a transition predicate for A6 governed composition in the c = a + b corpus.
The key point is not terminology.
The key point is anti-laundering.
A system must not launder uncertainty into action.
A witness record after the fact does not make an unclassified transition lawful.
A missing claimant is not proof that the successor space is exhausted.
A checker is not a court.
A hidden judgment inside a function name is still judgment, not computation.
This addendum went through four b-layer semantic review passes. The review cycle did not continue indefinitely. It converged.
The sequence moved from a blocking issue in the core successor predicate, to a second-order predicate on SPLIT, to a minor leaf predicate, and finally to closure of the predicate-review surface.
That matters because governance documents often fail in two opposite ways:
They either remain too vague to implement, or they pretend every ambiguity can be resolved mechanically.
This document takes the middle path:
make decidable predicates explicit; mark windowed predicates as time-state records; route escalate-only predicates to review; default unresolved transitions to HOLD.
Publication page: https://ivankotov.eu/publications/a6-ctp-v0-1-4/
DOI: https://zenodo.org/records/21227129
Parent A6 artifact: https://zenodo.org/records/20752182
This is not a product claim, not legal advice, not a safety proof, and not a claim of personhood.
It is a bounded governance artifact: a way to say, precisely, when a composition continues, exits, splits, holds, or ends.