The claim
A36 §2D had said that the cardinal residual of A35 is not defined “if each process supplies only an ordinal ranking over outcomes”. The owner, at U37: “I think this is why some parts of our research have leaned on the machinery of posets, lattices, presheafs, etc. We can do quite a lot with access to an ordering, even if it doesn’t give us a true measure.” A37: “I agree with your correction. My earlier discussion treated the absence of cardinal utilities as primarily a limitation. But for much of our argument, an ordering is sufficient, and sometimes it is the more appropriate structure.”
The construction (A37 §4, corrected). Let 𝒪 be the organization, ℛ(𝒪) the outcomes or histories its admissible policies can produce, and A_i the outcomes that meet demand i’s requirement. Demands are conditions on outcomes; what is attainable belongs to the organization. So the same demands can be compared when the organization’s policy space changes (CLAIM-361 is the claim that relies on this).
How A_i is specified depends on what the demand is.
- An improvement demand on a reference y₀ is A_i = {y : y ≻_i y₀}. Strict improvement may be impossible at an optimum, and then the demand is unmet there.
- A maintenance demand, “do not let this fall below its present level”, is A_i = {y : y ⪰_i y₀}. The reference meets it, which is what a successfully maintained requirement should do. If every demand in a collection is of this form, the baseline meeting all of them is not a defect. It means those requirements are compatible.
- A threshold or obligation, A_i = {y : y ⪰_i t_i} or a set given directly, may or may not hold the reference.
A preference is not a demand. That a more preferred outcome exists does not show that the present one fails a requirement.
The construction was corrected twice in review of the filing (workbench entries of 2026-10-18). The defect in A37’s version, A_i = {u : u ⪰_i u₀} “(or its strict version)”, was that it represented demands for improvement by weak-preference sets. Those cannot exclude the unchanged baseline, so improvement demands came out trivially compatible. The first correction overstated this, as if a weak-preference set could never be a demand. That is wrong for maintenance demands, as above. Weak Pareto improvement on y₀ (every demand ⪰_i y₀, at least one ≻_i y₀) is a condition on the collection, not an intersection of per-demand sets, so it falls outside this construction. The worked examples below are improvement demands with strict or threshold sets.
A set S of demands is jointly realizable when ℛ(𝒪) ∩ ⋂_{i∈S} A_i ≠ ∅. That is a fact about what can be attained. Whether the outcome the organization actually produces meets them is a separate fact. The compatibility complex is the first:
𝒦(𝒪; D) = {S ⊆ D : ℛ(𝒪) ∩ ⋂_{i∈S} A_i ≠ ∅},
with the empty intersection read as ℛ(𝒪). Its minimal non-faces are minimal incompatibilities. The incompatibility belongs to the collection, not to any one member. If A₁ = {a}, A₂ = {b}, a ≠ b, and the organization can attain either but not both, each demand is individually realizable and the pair is not; which one goes unmet depends on what the organization does. In the local model of a learner the attainable set is the set of admissible updates d, and the improvement demands are {d : g_iᵀd < 0} or {d : −g_iᵀd ≥ ε_i}, ε_i > 0. The facts below are stated with 𝒦 written as {S : ⋂_S A_i ≠ ∅}, the A_i read as the attainable points that meet each demand.
Structural frustration is defined by joint infeasibility. Given an organization 𝒪, a reference collection of demands D and its compatibility complex 𝒦(𝒪; D), a collection S ⊆ D is structurally frustrated exactly when S ∉ 𝒦(𝒪; D): no outcome attainable under the specified constraints satisfies every demand in S simultaneously. With a horizon H and a budget B made explicit:
F_struct(𝒪, S; H, B) iff ℛ(𝒪; H, B) ∩ ⋂_{i∈S} A_i = ∅.
This is a relational property of a collection of demands, relative to an organization, a horizon and an admissible set of transformations. It does not require cardinal utilities, experienced distress, or an identifiable process actively interfering with another. Minimal non-faces of 𝒦 identify the minimal structurally frustrated collections.
The formal criterion is settled by the stated definitions. What remains open is its broader organizational interpretation (QUESTION-131): when structural incompatibility shows up as operative resistance or thwarting, whether operative thwarting can occur without structural incompatibility, and what evidence is needed to attribute either condition to a particular organizational bearer. In particular, merely failing to realize a jointly attainable set of requirements establishes neither structural frustration nor operative thwarting.
Relief (A37 §4C). Relief of structural frustration is strict enlargement of the compatibility complex under fixed reference demands and acceptance conditions: 𝒦_before ⊊ 𝒦_after, so that at least one previously frustrated collection becomes jointly realizable and no previously compatible collection is lost. A37: “If the system simply deletes a difficult demand, it has changed the comparison problem rather than shown an expansion of its ability to realize the original demands.” And: “two reorganizations may relieve different conflicts, with neither compatibility complex containing the other. They would be incomparable rather than ranked by a universal success metric.” Suppressing a demand’s evidence, or lowering its weight, leaves its A_i where it was, so it leaves 𝒦 unchanged; it is not relief (A35 §4: “suppression, not necessarily resolution”).
The claim is that this is the basic form of frustration and of its relief, and that scalar measures are refinements of it (CLAIM-365), not definitions.
The definitional facts
These follow from the definitions and are checked here.
- 𝒦 is downward closed. If S ∈ 𝒦 and T ⊆ S, then ⋂_T A_i ⊇ ⋂_S A_i ≠ ∅. So, when U ≠ ∅, 𝒦 is an abstract simplicial complex on the individually satisfiable demands.
- Its minimal non-faces are the minimal conflicts. A set S ∉ 𝒦 all of whose proper subsets are in 𝒦 is a collection that is jointly unsatisfiable while no smaller collection explains it. By downward closure, every jointly unsatisfiable set contains one.
- Pairwise compatibility does not give joint compatibility. A37’s example: three binary variables with the demands a = b, b = c, c ≠ a. Each pair is satisfiable, the triple is not. The same holds for local improvement demands on a learner, which is the point G5 worked and A33–A37 did not make. Take three unit gradients at 120° in a plane, g₁, g₂, g₃ with g₁ + g₂ + g₃ = 0, and A_i = {d : g_iᵀd < 0}. Any two are not antiparallel, so −(g_i + g_j) lowers both: g_iᵀ(−g_i − g_j) = −1 + ½ < 0. All three cannot be lowered at once, since Σ g_iᵀd = 0 for every d. So 𝒦 is the boundary of a triangle, and the minimal conflict has three members.
- Pairwise non-conflict does rule frustration out. If every g_iᵀg_j ≥ 0 and no g_i is zero, then |Σλ_i g_i|² ≥ Σλ_i²|g_i|² > 0 for every λ in the simplex, so 0 ∉ conv{g_i} and a common descent direction exists (CLAIM-366). With nonzero demands, a minimal conflict of two is an antiparallel pair. Among non-antiparallel demands the minimal obstruction needs three, which is CLAIM-153’s “the minimal model of the obstruction needs three” in the update setting. For convex acceptance sets in ℝᵖ, Helly’s theorem bounds it above by p + 1 (CLAIM-365).
The claim’s interpretive part is the rest: that these sets are the right representation of an organization’s demands, that “relief” should mean enlargement for fixed demands, and that the comparison should be left partial.
Later versions of the same construction
- Indexed to an organization, horizon and budget. A38 §2: 𝒦(𝒪_t; D₀, H, B) = {S ⊆ D₀ : ℛ(𝒪_t; H, B) ∩ ⋂_{i∈S} A_i ≠ ∅}, where ℛ is what organization 𝒪_t can attain within horizon H and budget B. In A38’s words, the possibilities “belong to a particular organization with a particular developmental history”. Fixing D₀ across time, and saying what happens to a demand when the organization changes, is CLAIM-361’s, not this claim’s.
- Robust to compromised constituents. A41 §5 defines 𝒦_f^rob(𝒪), the sets S for which one admissible policy meets every demand in S against every allowed insider behaviour. It is the same complex with a ∃∀ condition in place of ∃.
- A flexibility order. A42 orders governance policies by inclusion of the contexts in which they respond acceptably: π₁ ⪰_flex π₂ iff ℛ(π₂) ⊆ ℛ(π₁). It is the same inclusion construction, applied to contexts instead of demands.
Relations
- CLAIM-128 and TERM-043. Translations whose distortion profiles trade off are incomparable until a task fixes weights. This claim is the same move for organizations: a partial order by inclusion, a scalar only once weights are fixed.
- CLAIM-153. The contextual version has torsors and a cycle class, which plain set incompatibility does not supply. A39: “ordinary set incompatibility doesn’t automatically supply the torsors, transition maps, or cohomological obstruction used there.”
- THEORY-232. Watson compares the parts’ total frustration under two organizations. Inclusion of complexes refines that comparison: it does not credit an organization that lowers the total by sacrificing a demand, and it leaves two organizations unranked when each relieves a conflict the other does not.
- A37 §5 on Arrow. The unanimity relation ⋂⪰_i of preorders is a preorder, so the organization’s evaluative state can be kept partial and selection of one action treated as a separate governance step.
What it does not say
- That cardinal measures are useless. A37: “We need cardinal quantities only when we want to distinguish how far one outcome is from another, how much improvement is required, or what resources a transformation costs.”
- That a partial order chooses. A37: weakening the output to a partial order “does not eliminate the practical problem of selecting an action when alternatives remain incomparable”.
- That demands have acceptance conditions. A37: “It does require each demand to have a well-defined acceptance condition.” Where they have none, the construction does not apply, which is the defeater.
- That every revision of a demand is a failure. Constitutive revision may be legitimate transformation (A38). It is simply not relief of the original demands.
Note of 2026-10-29: promoted
Promoted on CLAIM-368 and TERM-060, both Active, with no objection filed. Watson’s comparison (THEORY-232) is the measure this claim refines, cited as corroboration, not as a ground. CLAIM-449 objects to CLAIM-440’s use of the complex, not to this construction.