Why it is a question
A10 §10, Question 1: “Can heterogeneous operators spontaneously develop stable, adaptive patterns of mutually regulated influence?” A7’s ideal first result: “A population of initially undifferentiated, heterogeneous operators develops specialized niches, persistent but revisable trust, and higher-level governance structures solely through local transformation, temporally discounted policy evaluation, and constrained endogenous endorsement.” A7 §11’s H1 and H7 are its two halves. The owner’s condition at U5 was that niches would emerge “if permitted to evolve appropriately”.
The record holds two theses that answer it in advance, CLAIM-277 and CLAIM-279, and neither has a model behind it. The design constraint is A6’s: “The crucial experimental design choice is not to hard-code these outcomes.” CLAIM-288 limits what a positive answer would show. It would show that governance emerges from fixed local objectives, not that the organization constitutes itself.
An answer is a population model with no governance or role primitives, the parameter regimes in which durable differentiated authority forms or fails to, and ablations showing which mechanism each regime needs (A7 §12.4).
Note of 2026-10-18: the owner’s niche idea again, and a sharper experiment
The owner returned to niche specialization at U46 of the exchange of 2026-10-17: “this looks like a potentially useful integration point for ideas from ecology like niche specialization.” It is the idea’s second appearance (CLAIM-279). A46 found this question and strengthened its experiment with ecological criteria:
- Compare allocation of influence “based on independent contribution estimates, similarity to the dominant policy, or historical endorsement alone”.
- Ask “whether specialists emerge spontaneously, whether cooperation reduces higher-order arbitration costs, whether a rare specialist can regain influence when its relevant context returns, and whether the organization can reorganize after losing one of its established specialists.”
- Compare against “ordinary mixtures of experts, contextual routing, and multitask-learning baselines. Those architectures already produce specialization.” The model must show something more, “such as endogenous revision of roles and authority under changing conditions”.
From Chesson (2000), A46 took the distinction an answer should report: coexistence through stabilizing niche differences, not equality of authority (“Equality of authority ≠ coexistence through differentiated competence”). The failure mode an answer must rule out, niche exclusion by authority deprivation, is noted on CLAIM-277. Nothing has been run.
Note of 2026-10-18 (later): a formal bridge, and four sharper questions
A synthesis the owner relayed later on 2026-10-18 (workbench entry of that date) brought statistical feature learning and ecological coexistence theory to this question.
- The potential case is exact. In symmetric resource-mediated competition, A = UᵀU and r = Uᵀb − c, the population dynamics are multiplicative gradient flow on a convex feature-selection objective. A rare component’s invasion fitness is the negative partial derivative of that objective, and uninvadable equilibria are its KKT points (CLAIM-369). So in that case the recovery test of the note above is a marginal-value test. The identity fails once interactions are asymmetric or the operators’ objectives compete, and that is where this question’s organizations live.
- Four refinements are filed as questions:
- when feature geometry predicts coexistence (QUESTION-135);
- when differentiation yields positive recovery when rare, rather than imposed or entrenched roles (QUESTION-132), with recovery exponents λ^resident and λ
- whether specialization can relieve structural frustration by enlarging the attainable set (QUESTION-133);
- whether influence-dependent training destroys a marginalized specialist’s competence (QUESTION-134).
- The proposed study is CASE-117. It sets ordinary mixture-of-experts routing as the baseline, and imposed against endogenous diversity as the central ablation.
Chesson (LIT-1219) is still unread. The synthesis recommends reading it before any negative-frequency-dependence criterion here is given stronger standing.
Note of 2026-10-28: the first part split off
The fourth status review split this question under ADR-049. Whether one update rule yields stable, differentiated authority without a role coded in, and whether it also entrenches, is QUESTION-295, Answered by CLAIM-277 on CASE-182. This question keeps niches, rare-context competence and mutual regulation.
Note of 2026-10-29: the ecology and mixture-of-experts works, read
Eight works the synthesis of 2026-10-18 brought to this question are now read. Two are positive instances with a qualification, and the rest limit what an answer can borrow.
- Positive, in a narrow setting. Cooper and West, LIT-1225 (NOTE-1369): division of labour emerges with no group-level selection and no coded roles, from kin selection on one shared rule, with relatedness and a single task. Power et al., LIT-1235 (NOTE-1368): differentiated organization emerges, but with a coded-in constraint, the normalizing trade-off, which the authors say the result depends on (Appendix H) and which is imposed, not derived. CLAIM-288’s caveat, that this is organization from fixed local objectives and not self-constitution, applies to both.
- History dependence, shown less than read. Bunin, LIT-1234 (NOTE-1397), derives where the unique outcome ends. That the alternative outcomes then depend on order of arrival is shown by simulation only, and “entrenchment of whatever gained influence early” is the synthesis’s reading of that phase, which the paper does not test.
- Assembly end-points. Hofbauer and Schreiber, LIT-1233 (NOTE-1379): the −i communities are assembly end-points, so whether a role can be filled depends on the end-points reached without it, as the synthesis read them.
- Revision needs removal or asymmetry. Marsland, Cui and Mehta, LIT-1230 (NOTE-1352): under symmetric resource competition assembly is monotone, and no assembly sequence can undo an invasion except by killing species.
- Where optimization ends. Blumenthal, Rocks and Mehta, LIT-1236 (NOTE-1366): loss of reciprocity produces chaos only when the correlation ρ falls below √(S*/M*), so the limit of reading such a population as an optimizer is a boundary that depends on packing, not a point.
- The bridge to Chesson is two by two. Letten, Ke and Fukami, LIT-1237 (NOTE-1377), translate Chesson’s terms into resource impacts for two species and two resources at a stable equilibrium. For many operators, Chesson’s pairwise ρ does not by itself say whether the community coexists.
- Rare contexts are not modelled. Li et al., LIT-1411 (NOTE-1380), prove specialization among identically initialized experts, but their tasks arrive uniformly, so whether a specialist for a rare context survives is not addressed.
Chesson, LIT-1219, is now Unavailable, so the note of 2026-10-18’s recommendation to read it first stays unmet. The question is unchanged and stays Open.
Note of 2026-10-29: order of arrival in the potential case, a one-way core, and Chesson again
- Serván and Allesina, LIT-1824 (NOTE-1422). In symmetric competitive Lotka–Volterra assembly with a positive definite interaction matrix, as a Gram matrix of independent feature vectors is, there is no entrenchment by order of arrival: every order of arrival ends in the same community, and no invasion sequence returns to an earlier one. Order-dependence needs a symmetric matrix that is not positive definite, which a Gram matrix cannot be, or asymmetry. That agrees with Marsland, Cui and Mehta’s monotone assembly, in the earlier note of this date, and sharpens it: in this class an early entrant can be displaced, but the community it is displaced into is fixed by the pool, not by history. The invasion graph of any symmetric system is acyclic (THEORY-1168).
- Ito and Kaneko, LIT-1842 (NOTE-1414). A population of identical processes under one update rule, with no governance primitive, in which differentiated and durable influence forms. That is the part QUESTION-295 (Answered) already holds. It shows nothing on niches or on mutual regulation: the periphery has almost no influence on the core (2 of about 1,344 possible links), so the regulation is one-way, and nothing in the model is a competence that shows in rare contexts.
- Chesson. His 1990 paper, LIT-1826, is Unfetched and his 1994 paper, LIT-1827, Unavailable, so the note of 2026-10-18’s recommendation to read Chesson before a negative-frequency-dependence criterion gets stronger standing is still unmet.
The question is unchanged and stays Open.