Recursion Resolution
The problem
The human should be the controller, but the human's capacity to control is itself state-dependent. A degraded human cannot effectively self-regulate. If the machine becomes controller, a new principal-agent problem emerges. If the human is left alone, the original problem persists.
The resolution
The machine is not the controller. It is the state estimator and low-level stabilizer. It maintains the neurobiological conditions under which the human can function as their own controller. The human decides direction. The machine maintains preconditions for those decisions.
Derivation
The catecholamine-PFC dynamics (Arnsten 2009, 2015) establish that self-regulation capacity degrades under stress. The controllability circuit (Maier & Seligman 2016) establishes that prior experience with controllable outcomes provides proactive resilience. Together these mechanisms create the recursion: the person needs self-regulation to manage stress, but stress degrades the capacity for self-regulation.
The resolution follows from the control topology (D1): the machine operates at a different level than the human. The human operates at the level of values, goals, direction. The machine operates at the level of neurobiological state maintenance. These are different loops operating at different timescales. The machine does not decide where the aircraft goes. It maintains the conditions under which the pilot can decide where the aircraft goes.
Analogy
Kalman filter, not thermostat. The filter estimates where the aircraft is. The pilot decides where to go. A pilot without a state estimator in instrument conditions cannot fly. But the filter doesn't fly the plane.
Engineering implication
The system never assumes the controller role, even when the human is degraded. Instead, it reduces to more conservative actuator classes (null, minimal state notice, micro-controllability) as estimated degradation increases. The state-conditioned gating table operationalizes this: as the regime gets worse, the allowed actuator set gets smaller, not larger. The machine does less, not more, as the human's capacity decreases.
This distinguishes the architecture from adaptive automation (machine takes over tasks when human is overloaded) and from JITAI systems (machine prescribes behavior changes). Both of those become controllers during degraded states.