Addendum 98: Enclosing Compatibility Occurs Before Microscopic Causality
The Exact Coin Proof, MaxX Statistical Dependence, and the EP Projection Error
Abstract
Statistical mechanics defines a macrostate through its compatible microstates. Yet microscopic descriptions are frequently interpreted as though the subordinate states and transitions independently cause the enclosing organization that defines their statistical meaning.
Recursively Realized Distinguishability (R²D) identifies a mathematical contradiction in this causal interpretation. Its finite coin construction demonstrates that the same unchanged subordinate multiplicity landscape can acquire opposite statistical directions under different enclosing compatibility relations. Neither the subordinate multiplicities nor the transitions available within their native landscape uniquely determine the directional organization of their enclosing realization.
The result is exact combinatorics. It requires no physical constants, thermodynamic laws, microscopic trajectories, or R²D occurrence-realization postulates.
Maximum entropy and maximum caliber, collectively MaxX, reproduce the resulting conditional statistics when supplied with the enclosing classification and compatible state domain. But the enclosing information is an input to the calculation, not an output generated by the subordinate distribution. MaxX cannot establish bottom-up causal sufficiency from conditional statistics that already presuppose the enclosing organization.
This is the distinction between a valid statistical projection and an explanatory projection (EP) error. MaxX is a legitimate mathematical representation of subordinate statistical organization. Assigning that projected organization causal priority over the enclosing relation required to define it reverses the mathematical dependence established by the count.
The coin construction therefore supplies a finite statistical proof of the failure of autonomous microscopic causal sufficiency within MaxX. R²D identifies the corresponding causal ordering: enclosing compatibility defines the form of possible realization; subordinate occurrences realize that form.
1. The Question Is Causal Priority, Not Statistical Fit
Suppose an enclosing system possesses an observable organization. Its subordinate components occupy states whose relative frequencies reproduce that organization.
Two explanations are possible.
In the first, subordinate components are assigned independent causal agency for producing the enclosing organization. Their statistical behavior is interpreted as the origin of the form.
In the second, the enclosing organization defines which subordinate arrangements are compatible with its realization. The subordinate components realize that organization without independently defining it.
Both descriptions may generate the same observable statistics.
Their numerical agreement does not establish equal causal sufficiency.
The decisive question is:
Can the statistical organization of subordinate states independently determine an enclosing form when the statistical calculation itself requires information specifying that form?
R²D’s coin construction supplies an exact answer.
The problem can be stated before introducing energy, temperature, force, probability, or any physical theory of molecular motion.
2. The Native Subordinate Landscape
At boundary \(B_n\), consider an ensemble of ten distinguishable binary peers.
Its macrostates are
\[ i=0,1,\ldots,10. \]
Their native multiplicities are
\[ \boxed{W_{n,i}=\binom{10}{i}.} \]
These multiplicities count compatible subordinate realizations of the respective macrostates.
They are possible counts. They are not observed occupancies or accumulated transition frequencies.
Consider the adjacent distinction
\[ 5\rightarrow6. \]
Its native multiplicity ratio is
\[ \frac{W_{n,6}}{W_{n,5}} = \frac{\binom{10}{6}}{\binom{10}{5}} = \frac56. \]
Therefore,
\[ \boxed{ \Delta S_{n,5\rightarrow6} = \ln\frac56<0. } \]
The native landscape provides greater multiplicity support for macrostate 5 than macrostate 6.
Nothing in this calculation specifies an enclosing state.
Nor does any subordinate transition create the multiplicity landscape. A transition changes realization within the already-defined domain. It does not create the number of possible realizations of that domain.
3. The Enclosing Boundary Defines Compatible Realization
Now consider ten equivalent \(B_n\) peer landscapes, each defined by the same ten-coin multiplicities.
Let their readable peer macrostates be
\[ (i_1,i_2,\ldots,i_{10}). \]
The enclosing boundary \(B_{n+1}\) classifies compatible joint peer states through
\[ \boxed{ J=\sum_{a=1}^{10}i_a. } \]
This classification defines a new enclosing macrostate.
It is not a restoration of one hundred individually readable coin states at \(B_{n+1}\). The objects entering the enclosing classification are the ten already-defined \(B_n\) peer landscapes.
Their compatible multiplicities compose through
\[ W_{n+1,J} = \sum_{\sum_a i_a=J} \prod_{a=1}^{10}W_{n,a,i_a}. \]
For the specified binary landscapes, the multivariate Vandermonde identity gives
\[ W_{n+1,J}=\binom{100}{J}. \]
The closed-form result preserves subordinate multiplicity consequences without restoring subordinate state identity as a primitive enclosing read.
Select one representative \(B_n\) peer with macrostate \(i\).
For enclosing macrostate \(J\), the remaining nine peer landscapes must satisfy
\[ \sum_{a\ne g}i_a=J-i. \]
Their compatible extension multiplicity is
\[ \boxed{ K_{n+1\mid n,g,i}^{(J)} = \binom{90}{J-i}. } \]
Consequently, the complete enclosing-compatible multiplicity associated with local state \(i\) is
\[ \boxed{ \Theta_{n+1\mid n,g,i}^{(J)} = \binom{10}{i}\binom{90}{J-i}. } \]
The first factor is native subordinate multiplicity.
The second is the compatible enclosing extension.
Neither factor represents observed occupancy.
The enclosing boundary defines their joint compatibility before a particular realized occurrence is counted.
4. The Exact Reversal Theorem
Compare local macrostates 5 and 6 under a fixed enclosing state \(J\).
The exact ratio of compatible multiplicities is
\[ \frac{\Theta_6^{(J)}}{\Theta_5^{(J)}} = \frac{\binom{10}{6}}{\binom{10}{5}} \frac{\binom{90}{J-6}}{\binom{90}{J-5}}. \]
Evaluating the two ratios gives
\[ \boxed{ \frac{\Theta_6^{(J)}}{\Theta_5^{(J)}} = \frac56\frac{J-5}{96-J}. } \]
The native factor is always \(5/6\).
The enclosing factor varies with \(J\).
Consider two enclosing states.
First enclosing state: \(J=50\).
\[ \frac{\Theta_6^{(50)}}{\Theta_5^{(50)}} = \frac56\frac{45}{46} = \frac{225}{276}<1. \]
The enclosing-compatible multiplicity favors macrostate 5.
Second enclosing state: \(J=60\).
\[ \frac{\Theta_6^{(60)}}{\Theta_5^{(60)}} = \frac56\frac{55}{36} = \frac{275}{216}>1. \]
The enclosing-compatible multiplicity favors macrostate 6.
The native subordinate landscape has not changed.
The local distinction has not changed.
The mathematical definition of native multiplicity has not changed.
Only the enclosing compatibility relation differs.
Therefore,
\[ \boxed{ \text{Changing only enclosing compatibility can reverse local statistical direction.} } \]
This is an exact finite-count theorem.
No realization postulate is required.
4.1 The reversal condition is also exact
The enclosing-compatible multiplicity favors macrostate 6 when
\[ \frac56\frac{J-5}{96-J}>1. \]
Within the domain where both compared states have positive compatible counts, this simplifies to
\[ 11J>601. \]
Because \(J\) is an integer, the direction changes between \(J=54\) and \(J=55\).
Thus the theorem does not merely establish that reversal is possible.
It determines precisely where the reversal occurs in this count construction.
5. Theorem — Native Microscopic Statistics Cannot Be Causally Sufficient
Consider the proposition that native subordinate statistical structure is independently sufficient to determine the directional support of the enclosing realization.
If that proposition were true, there would exist a rule
\[ D=\mathcal F(W_{n,0},W_{n,1},\ldots,W_{n,10}) \]
that determines the favored direction from the unchanged native landscape alone.
For the transition \(5\rightarrow6\), this rule must return the same answer whenever the native landscape remains the same.
But the exact coin construction gives
\[ D(J=50)<0 \]
and
\[ D(J=60)>0. \]
The native multiplicities supplied to \(\mathcal F\) are identical.
The directional outputs are opposite.
Therefore no single rule depending only on the native subordinate multiplicity landscape can determine both results.
\[ \boxed{ \text{Native subordinate statistical structure is not causally sufficient for enclosing-defined direction.} } \]
Proof: Identical native inputs cannot uniquely determine two opposite enclosing-compatible directional relationships. Yet exact combinatorics establishes both relationships. The omitted determining relation is enclosing compatibility.
This proof does not rely on calling R²D’s causal interpretation true by definition.
It establishes the statistical insufficiency of autonomous subordinate counting directly.
The subordinate events remain necessary realizations of compatible states. Their necessity does not give them independent causal priority over the state relation that determines which realizations count.
6. MaxX Reproduces the Result by Assuming the Enclosing Information
Maximum entropy provides an exact statistical representation of the same coin system.
Let the ten peer macrostate tuple be
\[ \mathbf i=(i_1,\ldots,i_{10}). \]
Supply MaxEnt with the reference distribution
\[ q(\mathbf i) = \frac{1}{2^{100}} \prod_{a=1}^{10}\binom{10}{i_a}. \]
This reference distribution already contains the native multiplicities of the ten peer landscapes.
Now impose the enclosing constraint
\[ \sum_a i_a=J. \]
Maximize relative entropy,
\[ H(p\mid q) = -\sum_{\mathbf i} p(\mathbf i)\ln\frac{p(\mathbf i)}{q(\mathbf i)}, \]
over distributions supported on that fixed-\(J\) domain.
The solution is the normalized conditional reference distribution:
\[ p^*(\mathbf i\mid J) = \frac{ \displaystyle\prod_{a=1}^{10}\binom{10}{i_a} }{ \displaystyle\binom{100}{J} } \]
for compatible tuples.
Marginalizing over the other nine peer macrostates gives
\[ \boxed{ p^*(i\mid J) = \frac{\binom{10}{i}\binom{90}{J-i}} {\binom{100}{J}}. } \]
Consequently,
\[ \boxed{ \frac{p^*(6\mid J)}{p^*(5\mid J)} = \frac56\frac{J-5}{96-J}. } \]
MaxEnt reproduces the exact enclosing-compatible statistical ratio.
But examine its premises.
Before calculating the distribution, MaxEnt required:
- The admissible peer-state domain.
- A reference measure containing their native multiplicities.
- The enclosing classification \(J\).
- The restriction that only tuples compatible with \(J\) participate.
The mathematical optimization did not generate the enclosing classification or compatibility relation.
It used them.
Thus MaxX reproduces the same statistical result by calculating the distribution of subordinate states given the enclosing form.
The successful calculation cannot then establish that the subordinate distribution independently caused the form on which the calculation was conditioned.
7. Theorem — Conditional MaxX Statistics Do Not Determine Their Enclosing Condition
This limitation can be proved without reference to the particular coin numbers.
Suppose MaxX supplies an entire family of conditional distributions
\[ p(i\mid J). \]
A joint distribution requires
\[ \boxed{ p(i,J)=p(i\mid J)p(J). } \]
The conditional family alone does not specify \(p(J)\).
Let \(\mu(J)\) be any normalized distribution over admissible enclosing states. Define
\[ p_\mu(i,J)=p(i\mid J)\mu(J). \]
Different choices of \(\mu\) generate different joint distributions while preserving exactly the same conditional family \(p(i\mid J)\).
Therefore,
\[ \boxed{ p(i\mid J)\ \not\Rightarrow\ p(J). } \]
The subordinate conditional statistics do not uniquely determine the enclosing statistical organization.
An additional reference measure, constraint, or law may be supplied to determine an enclosing distribution. But that constitutes additional mathematical information. It is not generated by the conditional MaxX result.
This establishes the relevant impossibility:
Within MaxX conditional statistics, the information required to define the enclosing form cannot be derived from the subordinate distribution that was calculated using that information.
The mathematical dependence cannot be reversed by interpreting the conditional result as a cause.
The same problem applies to maximum caliber when an inferred history distribution depends on a supplied reference path measure and specified history constraints. Optimizing over those histories does not independently generate the constraints under which their distribution was inferred.
8. Why Reconstructing the Whole From Its Parts Does Not Reverse Causality
An objection may be raised.
The enclosing state was written as
\[ J=\sum_a i_a. \]
Does this not demonstrate that subordinate states determine the enclosing state?
It demonstrates that the value of \(J\) can be calculated from a compatible peer-state tuple under the already-defined enclosing classification.
It does not establish that the subordinate tuple independently defines the classification.
The mathematical distinction is between:
\[ \text{evaluating an established classification} \]
and
\[ \text{defining the classification itself}. \]
At \(B_{n+1}\), the state \(J\) has meaning because the enclosing boundary defines which compatible peer arrangements belong to that macrostate.
Without that classification, an arrangement of peers has no intrinsic obligation to count as macrostate \(J\) rather than under a different enclosing distinction.
Nor does algebraic reconstruction of \(W_{n+1,J}\) imply that the enclosing observer has regained direct access to unreplaced subordinate coin states.
Recursive replacement preserves count consequences. It does not preserve subordinate identities as the same scientific objects at the enclosing boundary.
Consequently,
\[ \boxed{ \text{Computability of an enclosing count from compatible components} \ne \text{independent causal production of its defining classification}. } \]
This distinction applies equally to physical mechanisms. A mechanism can be sufficient to calculate or realize a particular transition once its enclosing conditions are specified. That does not make the mechanism the source of the conditions defining the transition’s significance.
9. The EP Projection Error
MaxX is a valid statistical projection.
It maps a specified state domain, reference distribution, and constraints into a distribution of compatible statistical realization:
\[ (\Omega,q,\mathcal C)\longrightarrow p^*. \]
The projected distribution may agree exactly with observation.
The explanatory projection error occurs when the result is given a causal role that reverses this mathematical construction.
The invalid causal account is:
\[ \text{subordinate statistical distribution} \longrightarrow \text{creation of its defining enclosing form}. \]
But the distribution was calculated only after the enclosing form had been specified.
The proposed cause therefore presupposes the object it is claimed to cause.
This is not a numerical defect in MaxX. It is a defect in assigning primitive causal agency to the projection.
R²D identifies the correct explanatory order:
\[ \boxed{ \begin{aligned} &\text{Enclosing boundary and classification}\\ &\Downarrow\\ &\text{Compatible subordinate possibility}\\ &\Downarrow\\ &\text{Subordinate realization and occupancy}. \end{aligned} } \]
The subordinate distribution is real.
Its statistical description may be exact.
Its successful representation does not make it the cause of its defining enclosing relation.
Thus:
\[ \boxed{ \text{MaxX is a valid projection.} } \]
But:
\[ \boxed{ \text{Bottom-up causal attribution to that projection is an EP projection error.} } \]
Invoking another physical theory to supply a microscopic causal law does not convert conditional entropy maximization into a proof of microscopic causal sufficiency. That would be a different argument using additional premises, not a conclusion supplied by MaxX statistics.
10. What the Proof Establishes Before Physical Mapping
The finite coin construction establishes three results independently of physical dimensionalization.
First: possibility precedes occupancy.
The multiplicity of a state is determined by its compatible realization domain. Observed occurrences select realized alternatives; they do not generate the cardinalities of the states they occupy.
Second: enclosing compatibility determines statistical directional support.
The native local landscape can remain unchanged while enclosing-compatible multiplicity reverses the direction favored by that landscape.
The exact reversal is a property of the count construction, not a fitted kinetic effect.
Third: conditional statistical inference cannot establish independent bottom-up causal sufficiency.
MaxX can infer subordinate distributions from supplied enclosing constraints. That operation cannot independently derive the enclosing organization it assumes.
These statements concern possible-count organization and the logical dependence of statistical inference.
The subsequent R²D relations connecting possibility, retention, occurrence, occupancy, and recurrence determine how that organization is realized. They do not have to be assumed to prove the compatibility reversal.
This keeps the mathematical proof distinct from its later physical applications.
11. Consequences for Scientific Explanation
The result changes how successful microscopic explanations must be interpreted.
In thermodynamics, molecular states realize distributions organized by enclosing thermal and chemical relations.
In muscle, myosin switching realizes a force-dependent structural and mechanical organization. The individual switch does not independently establish the ensemble’s force–velocity relation.
In biological organization, subordinate molecular events realize cellular, organismal, and ecological states whose compatibility conditions are not exhausted by those local events.
In gravitational and quantum descriptions, physical coordinates and state representations can be accurate reads of enclosing relations without becoming their primitive causal source.
These are domain-specific applications of the same statistical ordering. Their physical mappings must retain their own definitions and observational content.
But the foundational distinction does not change:
The causal agency assigned to a subordinate mechanism cannot be inferred merely from the success of a statistical or dynamical projection that already requires the enclosing organization as input.
This is particularly consequential for AI-assisted science.
An AI model may reproduce an observed scientific law by adjusting its state space, statistical reference, or constraints. Such success establishes predictive or inferential capability under the supplied model.
It does not demonstrate that the model has discovered the causal organization generating the phenomenon.
A foundationally adequate scientific explanation must identify where the organizing information enters and whether the claimed cause actually supplies that information.
12. Principle — Enclosing Compatibility Before Microscopic Causality
The coin construction proves that unchanged native subordinate multiplicities can acquire reversed directional statistical support through a change in enclosing compatibility.
No native-only statistical rule can supply both directions.
MaxX reproduces the corresponding distribution only after the enclosing count relation has been incorporated into its inference problem.
Therefore:
\[ \boxed{ \begin{gathered} \text{Enclosing compatibility defines statistical support.}\\[4pt] \text{Subordinate statistical realization does not define}\\ \text{the enclosing compatibility on which it depends.} \end{gathered} } \]
R²D makes this mathematical dependence explicit as a recursive causal architecture.
The enclosing boundary defines the vessel of compatible possibility. Subordinate occurrences realize that vessel’s organization.
The proof does not require a physical protagonist, a microscopic force, or a mechanism transmitting instructions downward from a larger object.
Causal priority is assigned to the defining count relation, not to a separate agent operating on its contents.
Conclusion
The exact coin theorem exposes a limitation of bottom-up causal attribution within MaxX statistics.
The same subordinate multiplicity landscape admits opposite statistical directions under different enclosing conditions. MaxX correctly reproduces both results, but only by receiving the enclosing information that distinguishes them.
The mathematical explanation is therefore not that independent subordinate statistics create the enclosing form.
It is that the enclosing form determines which subordinate possibilities are compatible with its realization.
The contents realize the vessel. They do not independently create the vessel that defines what their organization can be.
MaxX describes that realization. R²D supplies the boundary-defined causal ordering.
Mistaking the first for the second is the EP projection error.
Sources and Logical Status
- R²D Canon, Part I, Appendix B, especially B.5–B.12 and the final discussion of causal agency: exact native and enclosing-compatible coin counts and statistical-direction reversal.
- R²D Canon, Part I, §§2.2–2.6 and 7.16–7.17: boundary-indexed statehood, possibility before occupancy, and the distinction between local realization and enclosing causal organization.
- R²D Canon, Part II, Chapter 5: Loschmidt, statistical direction, and the failure of reversible subordinate dynamics to supply the statistical arrow independently.
- Addendum 85, Local Agency and Enclosing Dependence: local realization, enclosing organization, and causal sufficiency.
- Addendum 96, Local Laws and Global Fit: the distinction between a domain-specific derivation and a claim of universal causal priority.
- Addendum 97, Mathematical Structure Before Local Laws: admissible mathematical organization preceding particular physical descriptions.
Formal status: The coin reversal and the insufficiency of native-only statistical determination are exact finite-count results. The non-identifiability of an enclosing marginal distribution from subordinate conditional statistics is an exact probability-theoretic result. The EP-error classification follows when those statistical projections are assigned autonomous bottom-up causal priority. Physical realization laws are not premises of these results.