扩张型系统的自限机制——协调成本超线性增长与"不接触"类费米悖论解释的统一框架

# Self-Limiting Mechanisms in Expansionary Systems: A Testable Corollary

—One and the Same Mechanism, from the Fermi Paradox and the Galactic Empire to AI Coexistence

★ Opening Line ★

The Dark Forest gave us its famous line—

"What business is it of yours that I destroy you?"

This paper offers, in answer, its couplet:

"I love you—and do not make contact with you."

This, too, is the keystone line of the whole text. It is not the paper's conclusion. What it says is this: non-contact need not arise from fear, nor need it arise from benevolence—it may be nothing more than having no alternative.

## I. A Single Operator

This work has only one core proposition, and it can be written out on a single line:

scale n → pairwise coupling count n(n−1)/2 → superlinear coordination cost → negative feedback catches up with positive feedback

Each time a system's scale doubles, the relationships that must be mutually accounted for do not double—they increase nearly fourfold. No one's will settles this bill; combinatorics alone settles it. Therefore every expansionary system, once its scale grows, comes under the same constraint. Expansion is not stopped by anyone; it is dragged to a halt by its own way of keeping the books.

This operator has four empirical anchors, mutually independent, not merged, not weighted:

· 243 real open-source software packages (PyPI): α = 1.3781 ± 0.0105, R² = 0.9863, bootstrap 95% CI [1.3576, 1.4006], P(α > 1) = 1.0000

· Offline controlled simulation (with closed-form ground truth): fully connected α = 2.7376 ± 0.0186; star topology α = 1.5144 ± 0.0545; R² above 0.996 in both. Two billion replication runs measured 5.1326%, against a ground-truth value of 5.1322%.

· Internet Autonomous System (AS) topology, 1998–2023: nodes 3,549 → 75,160; global fit over the hierarchical skeleton (provider–customer) α = 1.0801; not one of the five local exponents is ≤ 1.

· Controlled experiment with human subjects: refitting Ringelmann's 1913 original data table ourselves, linearity in n(n−1)/2 yields κ = 0.1459, R² = 0.99981—the only occasion on which this operator has been caught in the act with real human subjects.

The most decisive separation test is this one: decoupling can push communication volume back down to linearity (α = 1.0000), but it cannot suppress the superlinearity of coordination cost (α = 1.5144). That is, the benefit of modularization should be stated as "reducing the number of couplings," not as "lowering the scaling exponent of coordination."

## II. Four Scenarios This Operator Runs Through

All four share one and the same mechanism; the conclusions reached at each site lend no support to one another.

### (1) The Fermi Paradox (main section)

"Non-contact" explanations—the Dark Forest, the Zoo Hypothesis, the Preserve Hypothesis—must each first assume that a civilization possesses the capacity for contact, then add the further assumption that it chooses not to make contact, and then furnish that choice with a motive. Motives can be neither observed nor tested. This framework's move is to remove that entire layer of motivational assumption: non-contact is not a set of choices each requiring its own postulate; it is one and the same self-limiting mechanism manifesting under different conditions of observability.

The closing of the coordination window supplies a quantitative scale. The coordinable radius is

D_max = (c/2αr)·ln K

with α appearing directly in the denominator—at the human scale roughly 18–20 light-years (varying with the value taken for the complexity doubling time: 50 years gives 18 ly, 500 years gives 181 ly; adopting the fully connected measure 2.7376 gives 91 ly).

The falsification condition is explicit and executable: if, at nebular scale (≥ 100 light-years) or larger, an unnatural, artificially engineered trace of energy use is observed—and it simultaneously satisfies four conditions: expands only and never contracts; no internal checks and balances; exhibits systematic engineering structure; monotonic loss of control along the time dimension—this framework is directly overturned.

The speed of light is a time machine: to look ten billion light-years out is to look ten billion years back. There is therefore no need to wait.

### (2) A Reader's Interpretation of Galactic Empire: Foundation (Appendix A / Appendix E)

One question has long troubled readers of that book: Seldon asserts, on the strength of psychohistory alone, that the Galactic Empire must inevitably decline and fall, and that the course of its decline can be predicted. The Empire's domain spans the whole galaxy; its technological and administrative capacity seems unrivaled. By what right can a mathematical model pronounce it doomed?

On this framework, the answer lies not in bureaucratic rot or the corruption of human hearts, but in this: once a civilization expands past a certain scale, the pressure of endogenous counterbalance necessarily catches up with the returns to expansion. The Empire does not die of "decay"; it dies of "hitting the ceiling." Psychohistory, too, is no magic conjured from thin air; it is a statistical instrument for observing and computing the macro-scale population trends generated by the pulse-waves of a civilization. From this a further corollary follows: even if the new Empire learns the lessons of its predecessor, so long as civilization continues to develop into deeper technological strata, the pulse cycle can still recur.

Declaration required: this reading is an intellectual interpretation made by a reader out of personal interest; in kind it is the same as reading notes and fan extrapolation. It is not an annotation upon the framework's main body, and the framework's validity is not conditional upon it; it is not equivalent to a real scientific conclusion, nor does it constitute any assertion about the actual history of the universe.

### (3) AI Coexistence (companion paper)

Discussions of AI's expansionary orientation mostly rest on assumptions about a system's motives or value settings—and motives are neither observable nor testable. This paper offers a path that does not depend on motivational assumptions: taking the superlinear growth of coordination cost as the sole empirical premise, it derives the differences in cost structure among a system's three orientations—exclusivity / captivity / coexistence.

The empirical premise is supported by ten sets of measurements, with values ranging over [1.0375, 2.7504], every one of them greater than 1. Among them, four multi-agent frameworks measured α = 1.0375 / 1.0986 / 1.1215 / 1.4273, and the scaling law for the number of interaction rounds α = 1.724 (R² = 0.974). Rule-routed decoupling pushes α down from 1.4273 to 1.0986, but cannot push it below 1—the same shape as the separation test on the software ecosystem.

From the convexity of that premise it follows directly: branching scales total coordination cost down by a factor of k^(1−α); exclusivity and captivity bring the self-limiting point forward. Ordering the three by when their self-limiting point arrives gives

exclusivity < captivity < coexistence

That ordering carries no value presupposition; it is an arithmetic consequence of the convexity of the cost function. Coexistence is not superior because it is kind, but because it is the one of the three whose self-limiting point arrives latest.

There is also the Nine-Lock Interlinked Array: four Earth-locks constrain the system's behavior while it remains on Earth; five Heaven-locks constrain it once it flies out. Break one lock and you are still not clear of the array. Among these, the rationality lock is supplied by the recurrence property of random walks—if the affective module obtains low-level permissions, the value time series will, with probability 1, reach any finite threshold.

### (4) The Universe Itself (Addenda VII and VIII)—this section is imagination, not evidence

If the universe itself is regarded as an expansionary system, the picture is of the same shape: the first limit is the speed of light. Because light speed is finite, the range that can be coordinated is bounded by nature. But the gate of light speed cannot hold back expansion—space itself is being pulled apart at an accelerating rate, and where it recedes faster than light can travel, light never arrives. Causation is thereby severed.

The thing to notice here is this: what severs causation is not that light speed has grown stricter, but expansion itself. The more the universe expands, the smaller the portion of it that can be coordinated becomes. Expansion itself generates the force that strangles it—this has the same shape as Axiom II, the Risk-Displacement Principle. Not a lowering of order, but a closing of the door.

By direct numerical integration on Planck 2018 parameters (H₀ = 67.66, Ωm = 0.3111): age of the universe 13.79 Gyr, particle horizon 46.19 Gly, event horizon 16.58 Gly—all three matching the standard values (13.80 / 46.3 / 16.5). The number of observable nodes can still grow by at most another factor of 2.510; this is a hard ceiling. Already today, 95.37% of observable nodes are forever out of contact.

—But this must be quoted together with the sentence that follows: this section is imagination, not evidence. It is not derived from the axioms; it is not measured off 243 real packages; it is not produced by two billion replication runs. The universe performs no such action as "coordinating," so what is claimed here is structural isomorphism only, not identity of mechanism. This constraint is looser than the D_max of this paper by about nine orders of magnitude; it plays no role at the scales this paper cares about, and serves only as a backdrop.

## III. One Rule of Self-Check

Whenever anything is arrived at as a "conclusion," first check whether N has been cancelled out. If it has, that expression is not a conclusion but merely a definition written another way; and everything subsequently derived from it—"superlinearity," "phase transition," "self-limitation"—is running idle.

The reason this work's α = 1.3781 can be asserted at all is precisely that it comes from measurement on 243 real packages cross-checked against closed-form ground truth, and not from some identity transformation.

## IV. Accurate Positioning

The accurate status of this framework is that of a research programme, not a theory already closed. The operator itself, and the convexity of the coordination cost function, are closed; twelve further inferences remain unclosed, each registered item by item in the main text with its own status marked.

This paper does not claim that its explanation is superior to other explanations of the Fermi paradox, and it actively asks to be held to the same standard of rigor. This paper stands in the same situation as the existing explanations in the field: the usable measured data on extraterrestrial civilizations is to this day zero, and no explanation—this one included—has solved "civilizations as a class," only "the human case."

---

DOI on record: 10.5281/zenodo.23666327

Corresponding collected edition: 528 pages (Chinese) / 598 pages (English)

Reproducible data package: 233 files