Definition
Fix a state S with a law, a family I of interventions on what governs which of a system’s capacities become effective (authority, access, precedence: TERM-023’s relations), and an outcome Y. The governance-response profile of a state s is the tuple of its response kernels,
J_I(s) = (K_i(· | s))_{i∈I}, K_i(y | s) = P(Y^(i) = y | S = s).
A15 §2 wrote it as J_I(s) = (P(Y_{t+1} | s, do(i))){i∈I}, “where (\mathcal I) is a specified family of interventions on authority, access, or efficacy-governing relations”. A coarse description T = f(S) is governance-response sufficient for (I, Y) when J_I factors through T, that is, when T(s) = T(s′) implies J_I(s) = J_I(s′). A17 §4.1: “A coarse representation (T=f(S)) is governance-response sufficient for ((\mathcal I,Y)) when (J{\mathcal I}) factors through (T).”
For a declared intervention distribution ρ, the deficiency Γ_ρ(T) = I(S; Y^(I) | T(S), I) measures the failure, and is zero exactly when the profile factors through T on the interventions ρ charges and the states of positive probability (CLAIM-407). For one ρ and one outcome it cannot rise under refinement (CLAIM-409).
Beside TERM-043, not under it. TERM-043’s causal component is “Differences in
responses to corresponding interventions”, Σ_a π(a) d(T_# P_o(Y | do(a)),
P_t(Y′ | do(τ(a)))). Version 1 of this entry read the profile as that
component with τ the identity, π the declared ρ and the interventions
restricted to the mechanisms of governance, and filed it as refining
TERM-043. That was wrong. The component compares the push-forward of a
population’s interventional law with a target’s law, a graded form of
Rubenstein et al.’s exact transformation (NOTE-1037). For the hidden
governance bit it is zero: marginalising K is exact, and even a
τ-abstraction, because the coarse model’s noise can carry K
(THEORY-848, THEORY-849), while the profile does not factor
through the coarse state and Γ = ρ(h = 1) bits. Sufficiency is the
state-conditional version: Beckers and Halpern’s compatibility condition
with the high-level context required to be the coarse state itself,
compared on one outcome (CLAIM-399). So the refines link was
removed on 2026-10-21. A16 §14 declined a new dimension: “I would be
reluctant to add a new term such as governance fidelity as an entirely
separate dimension without checking whether it is already a specific
instance of causal fidelity”. Checked, it is not an instance of that
component; whether the profile should become a further component of
TERM-043, or stay a criterion beside it, is not settled here.
CLAIM-122’s supervisory test, “intervene on the governance relations”, is
what the profile records.
What would settle it. Whether claims about will-organization that now say “behaviourally equivalent but differently governed” (CLAIM-122, QUESTION-086, QUESTION-092) are clearer stated in it, and whether the owner wants the will line to carry a formal term.
What it is not
- Not a measure of will, or of self-governance. A16: “It would not automatically be a judgment about self-governance.” It records the response to governance set from outside, P(Y | do(i)), not whether the system changes its governance itself (CLAIM-407’s “What it does not say”, and A27 §3: “An ordinary machine can respond dramatically to a technician rewiring it”).
- Not a new kind of fidelity. A16 and A17 both declined that.
- Not causal abstraction. Exact transformation, τ-abstraction and zero approximate-abstraction distance all accept the hidden-bit coarse-graining that sufficiency rejects (THEORY-848, THEORY-849, THEORY-843; CLAIM-399).
- Not identifiable from passive data. A16: “Ordinary passive data may not identify (\Gamma).”
- Not the behavioural profile. A description can be exactly closed for the ordinary dynamics and keep none of it (CASE-125).
Note of 2026-10-25: the matched-populations test, proposed three times
A38, A43 and A44 of the exchange of 2026-10-24, continued, proposed pairs of organizations “with matched collective interpretation and policy distributions but different internal governance”, exposed to the same intervention. That tests whether the collective description is governance-response sufficient in this term’s sense. None was run. A40’s and A41’s interventions on access change outcomes by construction (where operators differ, changing who acts where must change the output), and do not test it. The term is unchanged.
Note of 2026-11-02: proposed a fourth and fifth time
In the exchange “Assess SDT Coverage” (2026-11-02), A21 §9 (four matched systems) and A22 §10 proposed the matched-populations test again. Neither cited this term’s note of 2026-10-25, and it is still not run.