Consent as Modelling Capacity

Consent is not a signature. Consent is modelling capacity parity.

Valid consent requires three conditions simultaneously.

Comprehension of irreversible risk. The agent must understand what cannot be undone (Axiom R: records are irreversible).

Ability to model consequence. The agent must be able to compute (at least approximately) the downstream effects of the intervention on their own coupling capacity and the organism’s.

Freedom from coercive constraint. The agent’s decision must arise from their own configuration (AP02: sovereignty), not from external override of their resolution capacity.

You cannot consent to what you cannot understand.

When modelling capacity is absent or impaired (due to complexity, disease, age, or cognitive limitation), consent defaults to predictive substitution: the closest available model of what the agent would choose given (a) known prior preferences from their accumulated records and (b) the stabilisation metric.

This is not a guess. It is a computation from the record set the agent has already written.

Predictive substitution is structurally inferior to genuine consent because it introduces prediction uncertainty. The architecture therefore treats every case where consent is substituted as carrying higher threshold requirements for intervention.

If you cannot ask, you need stronger evidence before acting.

KS-33.3: If modelling capacity parity cannot be assessed — if there is no structural criterion for “sufficient comprehension” — the consent architecture is ungrounded.