Regime Table (State-Conditioned Gating)
Design hypothesis
This table is a design hypothesis, not a validated protocol. The specific regime boundaries, action assignments, and whether this decomposition is the right one are all empirical questions. The claim is structural: actuator selection must be state-conditioned, and there exist states where null intervention dominates.
Regime definitions
| Regime | Description | Allowed actuators | Prohibited actuators | Rationale |
|---|---|---|---|---|
| G0 (green-stable) | PFC online, evaluative governance functional | State notice, reflection, value alignment, observation surfacing | None except coercive rhetoric | User can process reflective and value-relevant content |
| G1 (green-fragile) | PFC functional but vulnerable, ascending limb near apex | State notice, micro-controllability, low-load reflection | Value conflict confrontation, high-load planning | Reduce cognitive load; avoid value conflict escalation |
| R0 (red) | PFC in reflexive mode, evaluative governance degraded | Null, minimal state notice, micro-controllability, autonomic tools | Value alignment, high-load planning, complex observation surfacing | Avoid uncontrollable demand; restore contingency |
| R1 (red-severe) | Severely degraded capacity, safety concerns | Null, minimal state notice only | All high-load prompts | Prioritize safety and load reduction |
| U (uncertain) | Cannot resolve regime with acceptable confidence | Null or minimal state notice | Any directive action | Uncertainty demands conservative action |
Clinical precedent
State-conditioned gating has established clinical precedent in motivational interviewing: balanced ambivalence exploration during unresolved ambivalence produces worse outcomes than selective evocation (Miller & Rose 2015; Magill et al. 2018).
Design note: productive stress
G1 may include both pre-degradation vulnerability and productive developmental stress. The system must not systematically intervene on productive stress. The U regime provides conservative protection. The understanding layer may eventually resolve this (see principle: productive-incoherence-distinction.md).
Discrete vs. continuous
An alternative is continuous confidence-weighted estimation with graduated actuator intensity. More consistent with the continuous inverted-U (Arnsten) and smooth MODAL gain curves (Inglis et al. 2022), but harder to test and harder to specify safe operating constraints. The discrete approximation is chosen for engineering tractability. E1 can answer "can we classify regimes?" more sharply than "can we estimate a continuous capacity variable with sufficient precision?"