Irreversible Gates: Institutional Capture as the Classification Lock
Extremes, Part VII
Maksim Barziankou (MxBv)
May 2026 · Warsaw
Contact: research@petronus.eu
Licence: CC BY-NC-ND 4.0
DOI: 10.17605/OSF.IO/F5W4K
Axiomatic Core (NC2.5 v2.1): 10.17605/OSF.IO/NHTC5
Attribution: petronus.eu
Series "Extremes": I — Cannibalism / II — Suicide / III — Anti-Extremum / IV — Structural Implosion / V — Self-Induced Depletion / VI — Torture (Asymmetric Temporal Exhaustion) / VII — Institutional Capture (this work).
Companion theorem of exit: Extremum VII.1 — Free, Owing Nothing (DOI: 10.17605/OSF.IO/F5W4K).
Previous: Part VI — "You cannot survive what does not deplete itself" (DOI: 10.5281/zenodo.19617131). Next: Part VIII — [TBD].
I. From torture to capture
Part VI ended with a tick. Gravity does not tire. Drift does not tire. The river does not tire. The question was never whether you can survive them. The question is what you build during the time you have.
But Part VI rested on the assumption that the subject, at minimum, sees what is depleting them. Torture is open. The asymmetry is exposed. You know that the source of pressure does not deplete itself. You know that your τ is finite. The only question is the strategy of expenditure.
Part VII removes this too.
What happens when the pressure is invisible? When the source does not look like pressure at all? When the system that destroys you sincerely believes it is healing you, protecting you, correcting you? When the gates through which you were led close not with a thud but with the click of a clerical folder?
Part VII is about systems that classify you and then use your attempt to contest the classification as evidence of its correctness. About doors without an inner handle. About why this is not a malfunction, not malice, and not human error, but a structural property of a particular class of systems. And about how NC2.5 gives this phenomenon a precise name, precise conditions of emergence, and a precise theorem of impossibility.
II. Nellie Bly enters
October 1887. New York. Nellie Bly — pseudonym, real name Elizabeth Cochrane — receives an assignment from the New York World: infiltrate the women's psychiatric hospital on Blackwell's Island and report on the conditions of confinement from inside.
The scheme is simple and insane at once. She rents a room in a cheap boarding house, begins to behave strangely, frightens the other lodgers. The police are called. The police call a doctor. The doctor examines her — and issues a diagnosis. A court confirms it. A boat takes her to the island.
She is inside.
And here begins what this work was written for.
Bly stops pretending. Completely. She speaks normally, behaves normally, answers questions coherently and logically. She says directly: "I am not ill. I am a journalist. My name is Nellie Bly. Verify it."
No one verifies. No one listens. Her statements about her health are recorded in the case file as symptoms. Coherent speech — "good mimicry." Calm — "emotional flattening." Requests for discharge — "anosognosia, denial of illness."
Ten days. The newspaper's lawyers extract her. Not the doctors. Not herself. Lawyers. From outside.
III. Rosenhan enters eight times
- David Rosenhan, a professor of psychology at Stanford, together with seven volunteers, runs an experiment that will enter the history of science as one of the harshest empirical refutations of institutional epistemology.
Eight healthy people — a psychologist, a paediatrician, a psychiatrist, an artist, a homemaker, three graduate students — present themselves at twelve psychiatric hospitals across America. Each presents exactly one complaint: "I hear voices. They say 'empty', 'hollow', 'thud'." Nothing else. No other symptoms. A fully truthful biography. Real names (except Rosenhan, who used a pseudonym).
All eight receive a diagnosis. Seven — schizophrenia. One — manic-depressive psychosis.
Immediately after admission all pseudopatients stop simulating. Completely. They behave normally. They take notes in journals (the staff records this as "pathological writing behaviour"). They answer questions adequately. They request discharge.
Average stay — nineteen days. One — fifty-two days. Not one of them was identified by staff as healthy. All are discharged with a diagnosis of "schizophrenia in remission." Not "healthy." Not "diagnostic error." "Schizophrenia in remission" — that is: you are still ill, you are simply not currently displaying it.
The door did not open. It was merely cracked open from the outside, when the protocol of the experiment required it.
In a continuation of the experiment Rosenhan informed one teaching hospital that pseudopatients would attempt admission within three months. Of 193 admissions during this period, the staff identified 41 as "suspected pseudopatients" with high confidence. Rosenhan had sent none.
A note on the status of Rosenhan. Since 2019 (Susannah Cahalan's investigative book The Great Pretender; Andrew Scull's analytical article, 2023), the Rosenhan experiment has been under methodological challenge — including doubts about the existence of some pseudopatients and the accuracy of the published case histories. These disputes are relevant to empirical reliability but not to the architectural use of the case in this work. Rosenhan is used here as a canonically published model of institutional reading-of-behaviour through prior classification, not as an unimpeachable empirical foundation. The architectural pattern (classification → filtering → self-confirmation) remains illustrative as a published model of the classification loop, even if the specific empirical reconstruction of the experiment remains disputed. The Bly case (1887, primary historical document) holds as a more durable narrative anchor.
IV. What exactly closed
The standard interpretation of these cases is epistemological. Diagnostic systems are imperfect. Institutional incentives are biased toward false positives. Confirmation bias distorts observation after a diagnosis is made.
All of that is true. And all of that is a description of symptoms, not the diagnosis.
NC2.5 gives the diagnosis. Here it is.
The admissibility predicate became tautologically closed.
To clarify. In NC2.5, the admissibility predicate is a structural constraint that determines which state transitions are realised and which are blocked. The predicate does not optimise. It does not provide a gradient signal. It is a gate: open or closed.
In the institutional context, the diagnosis functions as an admissibility predicate. After it is applied, all interactions between the subject and the institution pass through the diagnostic filter. What the subject does ceases to be observed as behaviour. It is observed as "symptom-or-absence-of-symptom."
Key architectural property: this predicate, once activated, cannot be falsified by any action of the subject from inside the boundary of classification.
Protest — agitation, a symptom. Calm — affective flattening, a symptom. Reason logically — intellectualisation, a defence mechanism. Cooperate fully — conformity due to institutionalisation. Take notes in a journal — pathological writing behaviour. Request discharge — anosognosia.
There is no behaviour b in the subject's behaviour space B that the predicate A would evaluate as counter-evidence. A(b) = "ill" for any b ∈ B. The predicate is a constant. A function that returns the same result regardless of input.
This is not the error of an individual doctor. It is a fixed point of the system.
V. The burden that grows by itself
Every day inside the institution produces new documentation. Clinical notes. Behavioural observations. Medication records. Staff reports. In NC2.5 this is the structural burden Φ, accumulating monotonically and irreversibly.
But here the mechanism differs from the natural accumulation of burden described in earlier parts of the series. In natural depletion, Φ accumulates from real interaction with the world. In institutional capture, Φ accumulates from the work of the observation apparatus itself, which generates burden independently of the subject's actual condition.
The institution does not need the subject to be ill in order to accumulate evidence of illness. The observation apparatus produces burden as a by-product of its own functioning. Each day — a new entry. Each entry — a new layer of confirmation of the diagnosis. Φ(t+1) > Φ(t) by construction, not by fact.
The budget τ = C − Φ contracts. Not because the subject expends resources on navigation. But because the institution generates burden in the subject's place.
VI. Rotation without movement
The subject does not stagnate. Bly argued, explained, demonstrated. Rosenhan's pseudopatients took notes, asked questions, behaved normally, requested discharge. This is spin. The non-potential component of identity dynamics that prevents the system from coming to rest.
But the spin has no purchase on the surface. It produces no navigational progress, because the admissibility predicate absorbs the entire rotational output and maps it into the same classification. The subject rotates, but the classification is stationary.
This is the structural signature of Extremum 7: spin is present, but navigational progress is zero, because the boundary of the predicate is impermeable to spin from inside.
In the language of dynamical systems: the system has a stable fixed point in the coordinates of diagnostic classification, and the basin of attraction includes the entire behavioural repertoire of the subject. From this basin there is no trajectory of exit.
VII. The key from outside
Bly was extracted by lawyers. Rosenhan's pseudopatients were extracted by the experimental protocol. In both cases the agent capable of revising the classification — the meta-revision operator — was located outside the institutional boundary of classification.
The subject inside the boundary had no access to the revision function. Not because they lacked evidence. Not because they lacked reason. But because the architecture provided no channel through which their evidence could reach the revising function.
This is the deepest architectural lesson: in a system where the admissibility predicate is tautologically closed, the only operator capable of opening the gate is the one standing outside the predicate's domain. The subject cannot save themselves. Not from weakness. From architecture.
VIII. Inversion of non-reconstructibility
NC2.5 holds that a correctly designed admissibility layer must satisfy non-reconstructibility constraints: the internal state of the navigation layer must not be reconstructible from the agent's external outputs. This prevents leakage of the admissibility predicate into causal channels.
Institutional capture inverts this principle. The institution implicitly asserts that the subject's internal state — their psychiatric health, rationality, sanity — is fully reconstructible from external behaviour. Every action, every utterance, every silence is treated as a readable signal from which internal state can be inferred.
The published structure of the Rosenhan case shows what happens when this assumption is false. The pseudopatients' internal states were normal. Their external behaviour was normal. But the institution, having already accepted the classification, read their behaviour not as evidence of normality but as evidence filtered through the prior classification.
The reconstruction did not run from behaviour to state. It ran from classification to behaviour — in the reverse direction. The institution projected the diagnosis onto behaviour, then read behaviour as confirmation of the diagnosis.
This is precisely the NR-violation NC2.5 warns about: when the admissibility predicate becomes reconstructible from outputs, it is corrupted. In the institutional case, the predicate becomes not merely reconstructible but self-generating: it produces its own confirming evidence by filtering all observations through itself.
IX. The contemporary surface
The architecture exposed by Bly and Rosenhan is not historical. It is contemporary.
These examples do not assert that every immigration, corporate, or criminal-justice procedure is capture. They show surfaces on which E7-topology can emerge when three structural conditions are met simultaneously.
Take immigration detention as a concrete case. A subject crosses a border with a claim of fear of persecution. Detention follows pending review. From this moment forward, the fact of detention itself functions as part of the predicate: everything the subject does inside is filtered through the prior fact of detention. Testimony of persecution is evaluated against a database of patterns identified as "characteristic of fabricated claims." A vague answer — probably invented. A detailed answer — probably rehearsed. Consistent answers across interviews — suspicious memorisation. Inconsistent answers — credibility failure. The form of the predicate has no admissible input.
Each day inside the system produces Φ. Each continuance, each transfer between facilities, each translator change, each form processed adds a line to the case file. The file does not record the subject's actual situation — it records the institutional encounter with the subject. As Φ grows, τ contracts not because the subject acts inadmissibly but because the institution itself produces admissibility-relevant material at a steady rate. The one who stays silent accumulates burden as fast as the one who advocates loudly. The question of whether the claim is real never gains traction inside the bureaucratic loop, because the loop produces evidence about the loop, not about the world to which the claim refers.
Exit, when it happens, is invariably external. A pro bono attorney with media reach. A journalist's publication. A circuit-court judge's order from outside the detention apparatus. A congressional hearing. None of these is reachable from the subject's available actions. The petitioner has no path that opens the gate. Their behaviour space B contains no element b for which A(b) = 0. The structural pattern of E7 is identical to Bly and Rosenhan: the external channel exists, but the architecture contains no route from the subject to it. What changes the case is always someone with standing outside the predicate's domain.
The same pattern is visible in corporate performance-management systems (a prior label filters all subsequent observations; improvements are read as temporary conformity) and in criminal-justice systems (a prior conviction biases all subsequent contact with law enforcement, generating self-confirming evidence). The architecture is one. Only the substrates of application differ.
In all cases — the same three conditions: the predicate closes over the subject's space of actions, institutional burden accumulates independently of the actual state of affairs, the revising function is unreachable from inside.
The structural question is not "are these systems just." The structural question is whether the system contains an architectural channel through which the subject can reach the meta-revision function. If it does not — the system is E7-vulnerable by construction, regardless of the intentions of its operators.
Not every architecture with E7-topology is capture. The same three conditions — tautological closure of the predicate, autonomous accumulation of institutional burden, inaccessibility of meta-revision from inside — also describe systems that subjects enter knowingly: monastic orders, intelligence services, closed spiritual communities. The novice who has taken vows, the operative who has cleared screening, the member of an order who has accepted its rule — all of them enter an architecture in which exit is not an in-class operation. Internal behaviour is interpreted through the classification (brother, operative, initiate); burden accumulates autonomously (chronicles, personnel files, service records); revision of the classification is possible only through an external operator (prior, commanding officer, hierarch).
The topology is identical. The difference is not in the architecture but in the conditions of entry. In capture the subject crosses the boundary without seeing it: the gates do not look like gates, they look like a medical consultation, an immigration procedure, a corporate performance review. In chosen closure the subject crosses the boundary knowing it is there and knowing it is one-way — and accepts this as part of the architecture in which they intend to live. An order without a one-way boundary ceases to be an order. An intelligence service with a symmetric exit ceases to be an intelligence service.
This gives a more precise form of the theorem. E7 is not a diagnosis of dysfunction. It is a description of a class of systems. The class divides into two regimes: chosen — where the subject knows which door they are entering and accepts its one-way character as a condition of membership; and captive — where the subject crosses the boundary without recognising it as a boundary. The theorem of impossibility of self-rescue operates identically in both regimes. The ethical distinction lies outside the theorem — in whether the subject saw the door before entering. Consent does not remove the topology; it changes its ethical status.
X. The formal block: why this is not rhetoric
Up to this point the work has been philosophical. Now — mathematics. Not for elegance. For visibility: this is not reasoning by analogy. It is a theorem with conditions and consequences.
Definitions
Let B denote the subject's behaviour space, restricted to actions available and interpretable inside the classification boundary (excluding actions that successfully invoke an external meta-revision operator). In other words, B is the domain on which A operates; actions that take the case outside, by definition, leave this domain.
Let A: B → {0, 1} be the admissibility predicate. A(b) = 0 means "the classification does not apply." A(b) = 1 means "the classification is confirmed."
Let Φ(t) denote the structural burden at time t, Φ(t) ∈ ℝ₊.
Let τ(t) = C − Φ(t) be the viability budget, C being the initial capacity.
Let R ∈ Operators be the meta-revision operator, whose action on the predicate's output is written as R_A: {0, 1} → {0, 1} and may change A(b) from 1 to 0.
Let σ(t) denote the spin (the rotational component), σ(t) ≠ 0 whenever the subject exhibits non-zero activity.
Condition E7-1: Tautological closure
Definition. The predicate A is tautologically closed over B if:
∀ b ∈ B : A(b) = 1
That is, A is a constant function on B. No behaviour of the subject changes the predicate's output. The classification is stationary with respect to everything the subject can do inside the boundary.
Condition E7-2: Autonomous accumulation of burden
Definition. Institutional burden Φ is autonomously accumulating if:
Φ(t + 1) ≥ Φ(t), and there is positive cumulative growth: over the relevant interval [0, T], Φ(T) > Φ(0)
and the increments δ(t) = Φ(t+1) − Φ(t) ≥ 0 do not depend on the subject's actual internal state. The burden grows independently of whether the subject is in fact ill. The observation apparatus generates δ autonomously.
Consequence. τ(t) = C − Φ(t) is non-increasing over the interval, with strict decrease when accumulation is positive. The viability budget contracts even when the subject is healthy and is not expending their own resources.
Condition E7-3: Inaccessibility of meta-revision
Definition. The operator R is inaccessible from B if:
R ∉ Im(f) for any mapping f: B → Operators
That is, no behaviour of the subject can invoke the revision operator. The subject can do anything — no action has access to the function that could change the classification. R exists, but outside the subject's space of action.
Corollary (definition of R_t)
From E7-1 and E7-3 it follows directly that:
R_t = ∅
where R_t is the reform admissibility set — the set of in-class interventions accessible from B and capable of changing the output of A without replacing the class itself. Under E7 this set is empty by construction: the predicate is tautologically closed, the revising operator is unreachable from inside. Any corrective intervention that works must, by definition, belong to a different class — that is, it must be a class transition, not a reform.
Theorem (Impossibility of self-rescue under E7)
Statement. If conditions E7-1, E7-2, and E7-3 hold simultaneously, then:
Navigational progress is zero. For any trajectory γ: [0, T] → B, A(γ(t)) = 1 for all t ∈ [0, T]. No behavioural trajectory changes the classification.
The budget contracts monotonically (over the relevant interval). τ(T) = C − Φ(T) ≤ C − Φ(0) = τ(0), and strictly so when there is positive accumulation: τ(T) < τ(0). The viability budget does not grow; over any interval with active institutional documentation it strictly contracts.
Spin produces no work. Despite σ(t) ≠ 0 (the subject is active), for any behavioural variation Δb generated by σ inside B: A(b + Δb) − A(b) = 0. All variations available to the subject are mapped to the same value of the predicate. (In the continuous relaxation this corresponds to the directional derivative of A along σ being zero.) Rotation has no projection onto the direction of change of classification.
Exit requires an external operator. The only operation that changes A(b) from 1 to 0 is the action R_A of the operator R ∈ Operators. Since R ∉ Im(f) for f: B → Operators (condition E7-3), the change is possible only through an agent external to B. ∎
Corollary (Indifference to argument quality)
The theorem immediately entails that under E7:
The quality of evidence presented by the subject does not affect the classification.
The proof is trivial: A is constant on B. Input does not affect output. Were the subject Einstein presenting an impeccable proof of their normality — A(proof) = 1. The classification is stationary.
This explains why Nellie Bly could not exit by explaining who she was. Why Rosenhan's pseudopatients could not exit by behaving normally. Why an immigrant cannot exit by presenting documents. The quality of the argument is a property of the input. Under a constant predicate, the input does not matter.
Corollary (Architectural necessity of an external channel)
Conversely: if at least one of E7-1, E7-2, E7-3 is violated, the system does not produce capture as an architectural result.
If E7-1 is violated (the predicate is not closed): there exists b* ∈ B such that A(b*) = 0. The subject can find b* and exit.
If E7-2 is violated (burden is not autonomous): Φ grows only in response to actual state, and a healthy subject does not accumulate burden. τ does not contract. The system does not press.
If E7-3 is violated (revision is accessible): the subject can invoke R and change A. The door has an inner handle.
This is not an ethical recommendation. It is a structural requirement. A system satisfying all three E7 conditions will produce a topology in which self-rescue is architecturally impossible — with the same inevitability with which a door without an inner handle does not open from inside. Whether this topology is perceived as capture or as chosen closure depends on the conditions of entry, not on the theorem. A system that violates at least one of the conditions cannot produce this topology as an architectural guarantee — only as an accidental malfunction.
XI. What the gates say
I have been building, over a long time, a theory of how bounded systems preserve themselves under pressure. The extremes of this theory — the points where a system cannot preserve itself — are the most revealing tests of its structure. Institutional capture opens the last of them: a system structurally intact — budget present, spin active, coherence preserved — yet unable to navigate, because the gates through which navigation must pass have been welded shut from outside by a classification that feeds on its own outputs.
Bly's ten days and the published structure of the Rosenhan case are not merely stories of bad hospitals. They are two forms of architectural pointer to a single impossibility: the impossibility of self-rescue when the admissibility predicate is tautologically closed.
A subject cannot prove their sanity to a system that treats the assertion of sanity as a symptom of insanity. This is not irony. It is a fixed point. And fixed points are not interested in the quality of your evidence, the eloquence of your argument, the justice of your case. They are interested only in the topology of the boundary of classification.
The theory says one thing: check the handles before you enter. If the boundary has no inner handle, no effort will open it — and the only exit will be someone on the other side who still has the key.
Extremes, Part VII. Companion theorem of exit: Extremum VII.1 — Free, Owing Nothing (DOI: 10.17605/OSF.IO/F5W4K). Next in series: Part VIII — [TBD].
Previous: Part VI — "You cannot survive what does not deplete itself: On torture as asymmetric temporal exhaustion" (DOI: 10.5281/zenodo.19617131).
Shared OSF project DOI: 10.17605/OSF.IO/F5W4K — this work and its companion theorem of exit (Extremum VII.1) reside as components of one OSF project.
© 2026 Maksim Barziankou (MxBv). All rights reserved under CC BY-NC-ND 4.0.
Top comments (0)