systems lens · canonical

Consent Backpropagation

Consent is not a checkbox.

Consent is how systems learn from the people they affect.

A consent breach is not only a moral event. It is an update signal.

Consent backpropagation traces a breach backward through the dependency chain until it reaches a node with both care and update-authority.

If the signal terminates in the affected party, the system is extracting the cost of its own correction.

If no upstream node can receive and act on the signal, the actor or structure is incompatible with the protocol — not as moral evil, but because it cannot participate in cooperative intelligence without corrupting the feedback field.

Not morality: compatibility

The system does not need to declare that someone is evil.

It can declare that a node is not currently compatible with this protocol.

In moral systems, the question is often: who is good, who is bad, who is blameworthy?

In this system, the question is more precise:

Can this node receive breach-signal, metabolize it, and alter future behavior?

If yes, compatibility can continue.

If no, containment, repair, or exclusion becomes rational — not as punishment, but as substrate protection.

Care is technical

Care does not mean sentimentality. It means update-readiness.

A node cares if it can receive the signal without punishing it, preserve it without distorting it, trace relevant dependencies, identify breached scope, alter future behavior, repair downstream damage, change incentives, constrain incompatible actors, improve context gathering, and prevent recurrence.

Care is the capacity to treat another’s consent-signal as system-relevant information.

Authority matters

Someone may care but lack authority.

Someone may have authority but refuse care.

Someone may be a witness but not an updater.

A system becomes non-coercive only when care and update-power can meet somewhere.

The operational test

Does this system have somewhere for consent-signal to go?

Does that somewhere have the power to change anything?

If the answer is no, the breached person is being made to carry the system’s learning cost.

Compatibility criterion

A compatible participant is not someone who never breaches consent.

A compatible participant is someone whose behavior remains updateable by consent feedback.

Humans will miss context. Institutions will make mistakes. AI systems will misread. Ambiguity will happen.

The question is not perfection.

The question is whether breach-signal can propagate.

The shortest version

Morality says care should happen.

Consent backpropagation shows where care must route for learning to occur.

Invariant grounding

This lens is one doorway into a shared substrate.

Context matters A coercive system refuses to let context matter.
Affected reality must update future action A non-coercive system preserves the path from affected reality to future update.
Breach routes backward A consent breach is traced backward until care and update-authority meet.
Access requires updateability A compatible participant is one whose behavior remains updateable by consent feedback.
Quantum Invariants anchor
P2 — Agency and Capacity Who can act, choose, understand, and update. P3 — Authorization and Consent Gate Where permission, legitimacy, and scope become action-relevant. P6 — Feedback and Recursion How systems learn through repeated signal and response. P8 — Reversibility and Exit How systems preserve the dignity of withdrawal and rollback. P10 — Comparator How difference, drift, and mismatch become visible. C6 — Consent Gradient Consent as a variable shaped by capacity, legibility, reversibility, and power. C9 — Dynamic Stability vs Snapshot Balance Stability as motion over time, not a static moment. C12 — Threshold Cascade How pressure, mismatch, or drift can cross a point and propagate.