nucleation

CLAIM-152: Any individuation built on sufficiency is relative to the target it is sufficient for: fix the outcomes differently and a different partition wins, so no boundary drawn this way is universal

6 min read

Properties11
StatusActive
Version6
Rolethesis
Warrantdemonstrated
Tagsindividuation, epistemology, mathematical-statistics
Linedistributed-agency, formal-apparatus
GroundsTHEORY-194 — On Monte Carlo samples of the 2D Ising model and of the 2D dimer model with added noise spins, a block coarse-graining trained to maximize mutual information with the system beyond a buffer recovered the variables renormalization already uses (Kadanoff block spins; the dimers' low-momentum electric fields, with no weight on the noise), and iterated on 128 × 128 Ising samples gave the flow away from the critical point, locating it to about 1% and giving ν ≈ 1.0 ± 0.15, while a restricted Boltzmann machine trained by contrastive divergence on the noisy dimer data coupled to the noise and to columnar patterns instead, THEORY-202 — In variational renormalization over general Boltzmann machines, neither a matched partition function nor Kadanoff's exact trace condition makes a coarse-graining a renormalization: the first holds for joint Hamiltonians whose marginal is not the data, the second for hidden variables that do not interact with the system, so the variables a renormalization keeps must be fixed by a further stated condition, THEORY-066 — A Markov-blanket partition is defined relative to a chosen internal set, so a graph has one around almost any set of nodes, and the formalism alone does not say which set is the system, THEORY-073 — In Friston's 2013 formulation and the applications the record holds, a Markov blanket does not individuate a system: where it falls is fixed by modelling choices made before it is traced (the coupling graph, the internal set and the time at which steady state is assumed), and a blanket read from a fixed graph is the same whether the system produced the boundary or it was imposed
SupportsCLAIM-157 — Relative to a stated target, a combined form is one system while its joint dynamics are nearly decomposable with the form as a block whose gap beats the Occam factor, and it lasts as long as that block persists; the unit is the gap-isolated block, not any member, CLAIM-348 — No coarse-graining is normatively adequate in general because it preserves stability, prediction or aggregate performance: a description adequate for an institution's performance can merge states its members' claims must distinguish, so adequacy is fixed by the evaluative target, and the target cannot be justified solely by the description that preserves it, CLAIM-363 — A normative judgment is well-defined only once its relata are fixed (the organization or standpoint it concerns, the standard, the scale and the context), and that dependence does not by itself make it relative: once they are fixed, the judgment can be false, CLAIM-367 — Whether a training-time regulatory condition belongs to a network depends on which unit is individuated for the training dynamics and whose jurisdiction governs the updates; in conventional training most of that governance falls to the optimizer–model–data system, so a condition of training is not thereby a condition of the network, CLAIM-401 — The conditions an effective description is commonly asked to meet are not of one kind: predictive sufficiency and approximation fidelity are adequacy relative to a declared target, metric and horizon; tractability restricts the operations under which adequacy is assessed; dynamical independence and memory closure are relations of the representation to the system, memory relative to a chosen projection; so a matrix over them as independent binary axes conflates kinds, CLAIM-405 — An individuation is well defined only once its criterion is fixed (the target it serves, the transformations its unity must survive, the scale), and that dependence does not make it arbitrary: once the criterion is fixed a verdict can be false, and differently fixed criteria can pick out units of one system whose boundaries cross rather than nest, none of them less real for it
ComplementsCLAIM-279 — Whether an intervention stabilizes or destabilizes a pattern is relative to a policy and a frame, so denoiser, critic and re-noiser are regimes an operator falls into, not kinds of operator, CLAIM-300 — A state sufficient for a family of decisions under a model makes the rest of an inquiry's history worthless for those decisions, and it can stop being sufficient when a source is discredited, the model is revised or the decisions change, CLAIM-350 — A coarse-graining preserves a deterministic judgment exactly when the judgment is constant on its fibres: the judgment then factors through the coarse description, uniquely on its image, and otherwise two states the judgment tells apart have been merged
UsesTERM-049 — combined form
Cited as “complements” byCLAIM-535 — An organization that acts on what it observes has a policy acceptable in every circumstance exactly when, for each observation, the circumstances it cannot tell apart share an acceptable action; so every circumstance having an acceptable action does not give a policy, a finer observation never removes one, and where acceptable actions overlap the coarsest observations that suffice are several and incomparable, so which distinctions must be made legible depends on which acceptable action the organization commits to

Graph View

The claim

From MOM-POS-001. A statistic T is sufficient for predicting Y. With no Y, the notion picks out nothing. So every individuation built on sufficiency, whether of an identity, a universal or a system, inherits a parameter it must fix, and fixing it makes the result relative to the choice. MOM’s case: sufficient for locomotion and metabolism, “organism” wins; for niche, “animal” does. Biological taxon and ecological niche cross-cut, so no partition is universal on both.

Why this record holds it, and what it does here

It holds by definition. Here it disciplines the boundary claims filed beside it:

  • The Occam-priced test (CLAIM-154) takes its candidate direction and prior from a model chosen for some purpose.
  • The persistence test (CLAIM-161) depends on which γ.

So a verdict that the dancing couple is one system (TERM-049) (CLAIM-175) is a verdict relative to a stated target, such as predicting the dance’s next beat. MOM-LIT-023 and this record’s LIT-603 make the same point about Markov blankets: a blanket individuates nothing until the variables are fixed.

What it does not say

  • That boundaries are arbitrary. Given a target, the criteria decide, and can be wrong.
  • That MOM’s further conclusion applies here. MOM argues this costs its source corpus its Platonism about universals. This record makes no such realist commitment, so the position is a caution here, not a defeater.

Note of 2026-10-12: evidence now linked

Four THEORYs now ground this claim:

  • THEORY-194: which variables a compression keeps is set by what it must stay informative about.
  • THEORY-202: the variables a renormalization keeps must be fixed by a stated target.
  • THEORY-066: the blanket formalism takes the internal set as input and never outputs it.
  • THEORY-073: where a blanket falls is fixed by modelling choices made before it is found.

The body already made THEORY-066’s point about Markov blankets through LIT-603.

Note of 2026-10-12: a competent-uptake standard on the pragmatic-transport line, left to the owner

PR #121 filed CLAIM-217 (Proposed), which takes as its standard of sufficiency “the uptake of competent participants who know the standings and norms in play”. That is close to the candidate this claim’s defeat condition names and calls circular, “what a competent user must predict”. TERM-040 version 2 and CLAIM-204 take up the same standard. CLAIM-217 grounds it in CLAIM-151’s partial constitution of the act by uptake. Neither entry says whether that grounding is an independent argument in this claim’s sense or the circular case, and the question is left to the owner.

Note of 2026-10-12: the interlocutor’s reading, as advice

The interlocutor’s reading of 2026-10-12 sides with this claim against CLAIM-217: choosing competent uptake “does not establish a uniquely privileged standard. It chooses a reference class and grounds that choice in a theory of social constitution.” It would keep competent uptake as a practice-relative criterion, which “preserves CLAIM-152 instead of defeating it”. Recorded as advice. CLAIM-213 files the reading’s middle position between this claim’s lack of a realist commitment and CLAIM-139’s ontological clause.

Note of 2026-10-14: a sufficiency with no target

The algorithmic sufficient statistic (THEORY-277, from LIT-239 and LIT-224, read in NOTE-787 and NOTE-776) is a sufficiency with no outcome variable: it is sufficient for the data string itself, relative only to the reference machine. That qualifies “any individuation built on sufficiency”. It is not a counterexample to the claim’s point. It still gives no unique boundary: at non-critical complexities, distinct and independent sets witness the same code length (for a random string, the prefix set and the suffix set), and the minimal sufficient statistic is defined only to logarithmic precision and cannot be computed.

Note of 2026-10-14: a formal instance

Sun et al. (LIT-235, read in NOTE-833; THEORY-320) give a formal instance. In their spectral analysis the same unlabelled data is partitioned by colour or by shape depending on which auxiliary labelled classes are present (Theorem 5.2 against Lemma 5.4). What the top eigenspace picks out is relative to what else the operator connects. Recorded as support; the claim is unchanged.

Note of 2026-10-20: the target is relative to an instrument

The causal states, the minimal sufficient partition for predicting a process’s future, are canonical only given the target. Crutchfield and Shalizi’s own logistic-map example (NOTE-978) adds that the target also depends on the measuring instrument. Under a generating partition the map has one causal state; under a non-generating partition it has countably many. THEORY-756 records that among fully predictive partitions the causal states are the coarsest. So sufficiency individuates relative to a target, and the target is relative to how the system is observed. That is this claim’s point, one level down.

Note of 2026-10-21: carried to effective descriptions

A1 and A2 of the exchange of 2026-10-20 proposed “task-relative sufficiency under transformation” and “the operational relativity of effective structure” as new organizing ideas, and A3 and A5 restated them as the thesis of a proposed paper. A2: “Effective descriptions possess multiple, logically distinct virtues, and no one of the examined virtues can be treated without qualification as a sufficient criterion for all the others”, and “effective structure is objective but not absolute”. They are this claim with CLAIM-300 and CLAIM-348, carried to effective descriptions, and TERM-043’s componentwise comparison carried from fidelity.

The thesis was not filed as a claim. Its parts were: the conditions are not of one kind (CLAIM-401), the leave-one-out that would test them (QUESTION-252), and what adequacy is relative to (TERM-065, the effective-description regime). A paper built on it would be a works question for the owner. A8 and A9 carried the same structure to individuation, filed as CLAIM-405, which rests on this claim.

Note of 2026-10-29: two instances, one with the target fixed and one with it chosen

Two readings of 2026-10-29 give instances.

  • Computational mechanics. Shalizi and Crutchfield, LIT-1126 (NOTE-1372), make the target explicit: the causal states are a minimal sufficient statistic for the semi-infinite future of the same observable. They are an instance of this claim with the target fixed by construction rather than chosen.
  • Semantic information. Kolchinsky and Wolpert, LIT-1149 (NOTE-1370), take viability as the target. It is a function the analyst picks, and the system’s boundary is an input to it. Their §VI then proposes to choose the boundary, the timescale and the initial distribution by maximizing semantic information, which is measured against that viability. A boundary found that way is relative to the chosen viability function. The paper also concedes that negative entropy does not track identity (“both a rhinoceros and a human have low entropy”), so “keeping existing” there means staying concentrated, not staying the same system.

The claim is unchanged.

Backlinks