Manufacturing the Reactive Subject

Fascomachy · Document Two

You will read this and think of someone. Notice, later, that it was not yourself.

Evidence tier: Falsifiable. Falsifiers stated for its main claims; not yet run. Evidence Protocol · Theme Explorer

Bottom line up front

  • The mechanism. Four primitives — isolation, epistemic destabilization, belonging pressure, affect induction — form a closed loop in which each stage removes the defence against the next. Isolation takes the external check. Destabilization takes the internal one. Belonging supplies the authority the destabilized subject now needs, and the cost of losing it re-isolates them. Affect is not a fourth stage. It is the clock.
  • The output. Not a belief set. A subject — a reorganized architecture of judgment, in which what counts as evidence, who may supply it, and what disagreement means have all changed. Doctrine sits downstream and is revisable, which is why formations of this kind reverse positions overnight without losing anybody.
  • The constraint. Three claims carry this and one is much weaker than the others. The order claim is the most load-bearing and the least falsifiable, no experiment can fix that, and it is admitted at the moment it is made rather than in a footnote.

This document cannot tell you whether a particular person is captured, and it should be read as worse rather than better if it ever appears to.


The year

Fiction. The program, the calendar and its keeper are invented. It illustrates the circuit; it is not evidence for it.

From a calendar retained after the program was dissolved.

In January they asked me to record how often I felt uncertain. The request seemed reasonable. Uncertainty was the problem, and measurement was how serious people proved a problem existed before assigning it to somebody with fewer permissions.

The form arrived every morning at 8:00. If I had not completed it by 8:15, a reminder asked whether avoidance might itself be evidence of uncertainty. By February I completed it at 7:58. The system congratulated me for improving my engagement.

In March the questions changed. Instead of asking whether I trusted the information, they asked which approved source I trusted most. There was no option for none. Selecting Other opened a text box limited to 50 characters and a meeting with my supervisor. I learned to choose the source whose logo appeared first.

In April we were paired with accountability partners. Mine was a man named Pell, who apologized before reporting me and reported me before I could apologize. We became friends in the administrative sense: each knew what the other feared and neither knew what the other wanted.

What he reported was that I had called the March change a change. I did not remember saying it. I still do not. It was not a serious allegation and it was almost certainly true — it is the sort of thing I would have said in April, to Pell, who apologized. But I could not produce the afternoon it happened in, and after a week of trying I stopped trying, and what I had instead of the afternoon was the report.

The review form required an example of a colleague’s constructive contribution. I gave Pell. It raised both our scores. I want to be exact about the sequence, because the sequence is the only part of this I can still defend: he reported me, I could not verify it, and then I named him as constructive — and all three were the correct use of the system, and the third was also gratitude.

In May the program introduced belonging scores. People with high scores were asked to mentor people with low scores. People with low scores were asked to explain why they resisted support. My score rose after I stopped explaining.

By June I could predict which answer produced which intervention. This was called literacy. I could also predict which silence produced which suspicion. This was called progress.

In July my mother asked why I checked my phone during supper. I told her the check-in closed at 8:00 and late entries affected the team. She said the team was not at the table.

The form required an example of external negativity. I had one.

I want it on the record that I hesitated. I hesitated for most of a Sunday. The hesitation is not in the record, because the record has a field for the example and no field for the Sunday. I entered it at 7:58 on the Monday. She was not named — the field was free text, and I described her as a family member expressing skepticism about workplace initiatives, which was accurate, and which is how I know that accuracy is not the safeguard people take it for.

In August Pell disappeared from the dashboard. We were told not to speculate. His previous reports remained part of my record, although his name had been removed for privacy. I was therefore accountable for a statement about my own conduct, made by a person whose existence the system no longer confirmed, which I had never been able to remember, and which I had answered by calling him constructive.

After that the months stop being reliable.

They reduced the number of reminders, because participants had internalized the practice. I felt proud when the first missing reminder caused panic. Nobody had instructed me to panic. Nobody had ever instructed me to panic — there was no field for it, no intervention that produced it, and no partner assigned to it. It arrived on its own and it did the work of every reminder they had stopped sending.

Some time in what would have been the autumn, I began correcting new participants before the facilitators could. I explained that the questions were not accusations, that support could feel uncomfortable, and that discomfort often indicated growth. Every sentence was true within the language supplied for saying it.

The program was suspended pending review. We were instructed to continue voluntary check-ins so the review would have usable data. Participation reached 100 per cent.

I know it was December because the form said so, on the last morning it arrived. I had stopped knowing otherwise some time before that.

Then no form arrived. At 7:58 I opened the old link. It returned an error. I refreshed until 8:15, then wrote my answers in a notebook so there would be a record if anyone asked.

Nobody asked. I completed the notebook for another year.

1 · A list is not a machine

Document One is additive. Read as intended it answers what was done to me. It cannot answer a harder question: why did that work, when the same thing tried on someone else produced an argument?

Belonging pressure applied to a well-connected person with intact judgment produces a disagreement over dinner. The identical pressure applied to someone two years into a narrowing world produces a convert. Same primitive. Different world.

An additive account has three ways to explain that gap and none of them survives the cases.

  • Individual difference — true and useless. It predicts susceptible people would be captured by any primitive in any order, which is not what the exit literature describes. It also flatters the reader, who is invited to locate susceptibility somewhere other than in themselves, and who has already done so, on the first line of this page.
  • Dose — right direction, wrong shape. The documented pattern is that each stage arrives with a reasonable justification and none of them looks like much. Sequential discrediting of friends is not intense. It is quiet, plausible, and cumulative, and it is the reason nobody notices a year until it is over.
  • Content — contradicted outright. Doctrine is famously incoherent in the formations that hold people hardest, and reversible at no cost to membership. If people were held by what they believed, changing what they believed would cost you the room.

What is missing is a composition mechanism — a specification of how primitives combine into something with properties none of them has alone. It does not fall out of the list. It has to be built, and then attacked.


2 · The circuit

[11] Isolation ─────▶ [2] Destabilization ─────▶ [9] Belonging
      ▲                                             │
      │                                             │
      └────── exit cost deepens isolation ─────────┘

[3] Affect — the clock
    compresses appraisal time and accelerates the loop

Stage 11 · Isolation and dependency

Taken here: the second opinion. Given back: simplicity. Reads as: focus.

You test your perceptions against other people. Remove the second opinion and the first becomes unfalsifiable — not wrong, unfalsifiable, which is worse and much quieter. Remove material independence as well (income, housing, documents, professional standing) and voice becomes unenforceable, because a person who cannot leave cannot credibly object, and both parties know it without either saying so.

This is the entry point of the whole apparatus and it is almost never experienced as isolation. It is experienced as a series of sensible decisions about difficult people.

She had been difficult. You were tired. Nobody did this to you.

Stage 2 · Epistemic destabilization

Taken here: your own record. Given back: humility. Reads as: growth.

With no external calibration left, sustained contradiction from a trusted source resolves more parsimoniously as I misremembered than as this person is lying to me on a schedule. The first is an ordinary thing that happens to everyone. The second is an accusation that would cost you the only relationship you have left, and you have already checked, and you have no one to check with.

The output is a subject who can no longer self-verify and therefore requires an external authority to supply reality. This is now a felt need rather than an imposition, and that inversion is the hinge of the entire circuit. Nothing further has to be forced.

You were certain, and you were wrong, and they were kind about it. Nobody did this to you.

Stage 9 · Belonging pressure

Taken here: the option of being wrong alone. Given back: a room on Thursdays. Reads as: home.

The group supplies exactly what stage 2 created the need for, in the most efficient available form: reality and membership in one package, from the same people, at the same time, for free. Because membership is now load-bearing, its withdrawal is catastrophic — and that cost re-enforces stage 11.

The loop closes here. It closes without anyone maintaining it. From this point the arrangement is self-sustaining and requires no further intent from anybody, which is why looking for the person responsible returns nothing and why not finding them is so disorienting.

They understood. Nobody had understood for a year. Nobody did this to you.

Stage 3 · Affect, which is the clock

Taken here: the interval. Given back: certainty. Reads as: conviction.

Affect is not a fourth stage. It runs the cycle below deliberation speed and supplies felt certainty in place of the verification that is no longer available. The taxonomy’s qualification carries here and matters: the primitive is unilaterally induced affect — supplied by the arrangement, timed to the ask — not strong feeling as such.

A person inside the circuit does not feel less. They feel more, on someone else’s schedule. Any account that treats passion as the contaminant has misread the machine and will build the wrong defence — and will also, incidentally, insult everybody it is trying to reach.

You have never cared about anything this much. It was the best year of your life. Nobody did this to you.


3 · The three claims, one of which is weak

Claim 1 — order matters, and this is the soft spot

The primitives are not commutative. Belonging pressure applied first produces argument; applied after isolation and destabilization, it produces conversion.

This is the most load-bearing claim in the document and the least falsifiable, and no experiment can fix that. You cannot run the circuit on people to see whether sequence matters. Nobody will approve it. Nobody should.

What is left is observational and weak: retrospective sequence from exit interviews, which is self-report about a period the subject has powerful reasons to reconstruct; natural variation in how formations recruit, which is uncontrolled; comparative case work, which is interpretive. That is the whole evidence base. It is stated here, at the claim, rather than in the falsification section — because a claim that announces its limits when it is made reads as confidence, and the same admission found later by a critic reads as concealment.

The weaker formulation — order plausibly matters — would be safer and would cost the document its spine. The circuit is the non-commutativity claim. Without it this is a list with an arrow drawn on it.

Claim 2 — the circuit is self-repairing

Each stage’s output is the next stage’s precondition, and stage 9 feeds back into stage 11. Perturbations get absorbed rather than propagated, which predicts which interventions fail and how. Present counter-evidence: it arrives as an attack from outside the group, which stage 9 predicts and stage 2 has already prepared the subject to discount. Restore one social contact: stage 9 raises the cost of using it. Reduce the affect: the loop runs slower. It does not stop.

Stable is not immortal. Kuran’s account of preference falsification supplies the other half of the record — visible consensus can detach entirely from private belief and hold until it collapses in weeks. A model that only explained stability would be wrong about a historical record full of formations that stood for years and then went overnight.

Claim 3 — the output is a subject

What changed is the architecture of judgment: what counts as evidence, who may supply it, what disagreement means, and what follows from being wrong. Doctrine is downstream of that architecture and revisable inside it.

The claim predicts that such formations reverse doctrinal positions overnight without losing adherents, because adherence was never to the doctrine. That is checkable against the historical record, and it is where this document is most exposed if the record declines to cooperate.

A note on the count

The companion argument identifies eleven safeguard families. This circuit has four stages. That mismatch is not a problem to be solved. Capture is a sequence; defence is a coverage set. A sequence has as many stages as its dependency chain has links. A coverage set has as many families as the counters group into. Different counts are what two genuinely different structures predict, and the mirror between these documents is multiplicativity, and multiplicativity only.

The mismatch has since been worked through row by row. The Hands on the Frame §2 maps all thirteen primitives against the eleven families, and reading the four circuit stages off that map produces a result this document could not reach alone. Of the four stages, only two are held by a family with nothing behind it: Records stands alone in front of stage 2, and Time stands alone in front of the clock that stage 3 runs. That makes Records and Time the pair the circuit cannot proceed without, while Exit and Voice — both multiply covered — decide whether the sequence closes into a loop or terminates. The falsifier is stated there: find a documented capture that ran to completion where the record was intact and deliberation time was uncompressed, and the pairing is wrong.


4 · From inside

Mechanism first, and the experience falls out of it — because a description of what it is like to be someone else, written by someone who is not them, is an essay, and this needs to be an argument.

High confidence does not merely accompany resistance to counter-evidence. It produces it. Rollwage and colleagues (2020) show that under high confidence, choice-consistent evidence is selectively accumulated while processing of disconfirmatory evidence is abolished — demonstrated behaviourally, modelled with drift-diffusion, confirmed with magnetoencephalography. Rollwage, Dolan and Fleming (2018) locate the population signature: holders of radical beliefs match moderates on task accuracy but are worse at recognizing their own errors and more confident when wrong — two general-population samples, n = 381 and n = 417, on a dot-discrimination task with no political content whatever.

Two further findings sit alongside it from different labs and methods. Dogmatic participants seek less corrective information under uncertainty and rely less on their own confidence signals to guide the search (Schulz et al. 2020). Dogmatism tracks slower evidence accumulation under drift-diffusion modelling across a 37-task battery (Zmigrod et al. 2021).

“Counter-evidence strengthens commitment” is not an observation in search of an explanation. It is what abolished disconfirmatory processing feels like from inside.

The counterintuitive core

The reactive subject does not experience deficit. They experience surplus.

From inside the loop it is more certainty, more purpose, more connection than before entry. Confusion, isolation and doubt are what the loop resolves. That single fact explains the pattern of failed interventions better than anything else on this page.

  • “Just show them the facts” fails because facts were never the binding constraint. The architecture that would have evaluated them is the thing that changed.
  • Counter-evidence strengthens rather than weakens — mechanically, because disconfirmatory processing is suppressed under high confidence, and socially, because it is legible as an attack from the out-group.
  • Exit is loss, not liberation. The subject is not escaping a deficit. They are giving up a surplus, and the exit literature reports exactly that, and anybody planning an intervention who has not absorbed it will fail kindly and completely.
  • Participants are not dupes but enthusiasts, and the error is consequential.

That last point requires meeting Haslam and Reicher’s engaged-followership reinterpretation of Milgram head-on rather than dodging it. People do not blindly obey; they actively identify with a cause and work creatively toward it. This complicates any account built on obedience — and it happens to support this one, because it moves the weight from authority to belonging, which is where this circuit puts it. The reactive subject is not following orders. They are solving problems, imaginatively and late into the evening, on behalf of a group whose reality they now depend on.


5 · Instantiations

Six cases, deliberately spanning scales, because a mechanism that appears at only one scale is probably an artifact of that scale’s description. One constraint throughout: real-world only. A fictional system specification demonstrates that a design is coherent, not that it works, and citing one as support hands a hostile reader a free kill.

CaseWhat it demonstrates
Coercive control (dyadic)The circuit at n = 1. Stark’s account maps position for position. The legal history is itself the evidence: incident-based law could not see it, and what made it legible was reframing the harm as a cumulative pattern of liberty deprivation — the additive-to-multiplicative move, made independently, in law, by people who had never heard of any of this.
The high-demand groupEscalating commitment does work the circuit does not require but benefits from. Treat the popular frameworks critically; use the primary exit research.
Platform-mediated radicalizationThe important variant, because nobody intends it. Engagement optimization supplies stage 3, recommendation systems supply stage 9, and isolation arrives as a side effect of displaced sociality. If the circuit needed a manipulator you could go and look for one. It does not, so you cannot.
1930s GermanyKlemperer on vocabulary as the delivery mechanism — stage 2 conducted through ordinary words, “like very small doses of arsenic.” Arendt on loneliness as precondition: stage 11 at population scale, achieved before the political achievement. A mechanism, not a comparison.
Freire’s antidialogical action (1968)Conquest, divide and rule, manipulation, cultural invasion — four characteristics mapping onto the circuit position for position, derived from colonial pedagogy with no access to cognitive psychology. Convergence is not proof. It is what a merely-invented structure does not usually do.
Pig butcheringThe full circuit at dyadic scale, for money, on a production line — plus staged demonstration, where the platform really does pay the first withdrawal. Nothing false is asserted. The inference is valid. The sample is the fabrication.

The amender

Fiction. The office, the figures and the village are invented. It illustrates the circuit; it is not evidence for it.

What the amender said over tea, once.

My name is Braided-Grain. I amend.

A figure is published. Later a better figure is found. I retrieve the first, amend it to the second and refile it. That is the work. There is no deception anywhere in it. Everyone knows the early figures are provisional, and we say so at the time, in writing, on the document I later amend.

I was good at it young. That is how I came to keep a book.

The book was mine. Nobody forbade it because nobody had thought of it. In it I wrote what a page had said before I changed it. Not from suspicion. From tidiness. A person likes to know the shape of a day’s work.

The trouble came slowly, and it came from the work being honest.

To record what a document said before amendment, I had to reproduce the superseded figure. The superseded figure is the one thing I am not permitted to reproduce, because reproduction is precisely what amendment exists to prevent. My book began to say only that an amendment had been made. Then only that some had been. Then it became a column of dates.

A column of dates on which I cannot remember what happened is not a record. It is an accusation, and the only person it accuses is me. I burned it. I was relieved. I told myself the relief was tidiness. I was thirty-one.

Years later a young man asked me over tea what the March figure had originally been.

He was preparing an application for winter grain. The district had received relief when the March figure first appeared, then lost part of the allocation when the amended figure showed the harvest had been larger. His village said the first distribution had already proved the need. The office said need must be calculated from the corrected harvest.

He wanted the original figure because the appeal form asked him to show what information the earlier decision had relied on.

Understand that I was pleased to be asked. I had a good memory. I was proud of it. I opened my mouth and heard my own voice give him the corrected figure in the confident tone of a person who had been there.

He showed me a copy of the first relief notice. It named the date but not the figure. Notices do not repeat provisional figures because repetition may preserve an error after correction.

I checked the amended page. The corrected figure stood there without scar or seam. That is how a proper amendment looks.

“Then the office was right,” the young man said.

I told him the record was clear.

His appeal failed. The winter allocation was calculated from the corrected harvest. By spring the village had borrowed grain against the next crop, which made the following harvest larger on paper because borrowed seed is counted when issued and debt is counted elsewhere.

Nobody lied to me. Nobody instructed me. I did every part correctly, one improvement at a time, and each improvement was an improvement.

When people later spoke of the poor March figure, I corrected them.

6 · Falsification

Two kinds of claim, needing two kinds of test.

Family A — output falsifiers. Does the reactive subject exist as a distinct state? The falsifier: if members of identified high-control formations show metacognitive sensitivity indistinguishable from matched controls on content-free tasks, the claim fails. That test has not been run. Existing work measures survey-assessed radicalism, not capture, and those are different things.

Family B — structural falsifiers. Is it a circuit? If populations with intact intermediate associations show the same susceptibility as atomized ones, the entry claim is wrong. If belonging pressure produces the same effect regardless of what preceded it, the circuit is a list after all. If people inside such formations reliably report doubt and constraint rather than clarity and purpose, the mechanism is misdescribed. If restoring durable external contact reliably fails to reduce commitment, the circuit is not load-bearing. If such formations lose adherents at the same rate as ordinary political groups when they reverse positions, Claim 3 is wrong.

The perturbation prediction, and its honest status. Level of agreement carries no information about how the agreement was produced. The proposed discriminator is response to perturbation: circuit-produced consensus should snap back in unison, deliberated consensus should scatter and re-converge by different routes. This is a research proposal, not an available instrument. No validated instrument administers a counter-argument and scores conclusion-holding against reason-movement, in any population. Two constraints bind any attempt to build one: reason-movement is the human default under coherence-based reasoning, so only its direction and cost could discriminate; and resisting persuasion raises certainty most in high-elaboration processors, so the probe hardens precisely the people the test would classify as reasoning well.

What the exit literature does and does not support. The defensible finding is decoupling, not temporal precedence: disengagement frequently occurs without deradicalization, driven by disillusionment with leadership, tactics and interpersonal conditions. The evidence is qualitative, retrospective and uncontrolled. So belief and behaviour are separable, and exit is predicted by satisfaction, investment and available alternatives rather than belief content — enough to predict why belief-first interventions underperform. It does not license “behaviour first, then attitude,” and an earlier draft of this document asserted exactly that, which a hostile reader would have dismantled on sight.


7 · What this cannot say

Its own section, not a caveat inside the last one, because a limit stated as an aside reads as throat-clearing and this limit is the load-bearing defence against the document’s largest risk.

What the evidence licenses. A manipulative apparatus produces a measurable population-level signature: reduced metacognitive sensitivity, higher confidence when wrong, on tasks with no ideological content.

What it forbids. That any individual can be identified as captured. The source authors state that metacognitive differences explained only a limited amount of the variance in radicalism. Limited variance explained is not a modest caveat. It is a prohibition on classification. Any classifier built on this would have unknown false-positive rates against the one error that matters most — misclassifying a well-grounded dissenter as captured.

  • The studies measured radicalism, not capture. A person may hold radical views through capture, through reasoning, or through having looked at the same world you did and reached somewhere else. The design does not separate them and neither can this document.
  • Direction is unestablished. Nothing shows whether reduced metacognitive sensitivity precedes entry or results from it. The circuit argument needs the second. The evidence supports neither.
  • The theory applies to its author and its reader. Anyone who finds this argument immediately and completely convincing should notice that this is precisely what the argument predicts they would feel.

A theory in which every disagreement proves the machine is unfalsifiable and contemptuous. The structural defence is not a promise of good faith, because a promise is exactly what a manipulator also offers. It is a refusal of the capability.

Which returns us to the first line of this page. You thought of someone. That was the instinct this section exists to disarm, and noticing that you had it is the only part of the exercise that was ever going to be difficult.


8 · Interruption

Stage 11 is the cheapest to break and the only one breakable from outside. That was a working hypothesis in earlier drafts. It now has a derivation.

Circuit stageFamily that counters itBreakable from outside?
11 IsolationExit — plural association, independent income, portable benefitsYes. Exit instruments are supplied by the environment, not by the subject.
2 DestabilizationRecords — archives, minutes, contemporaneous timestamped recordPartly. A record can be kept for someone; consulting it requires them.
9 BelongingSeparation — plural memberships, positions decoupled from personsNo. Plural belonging has to be practised by the subject.
3 AffectTime — latency, cooling-off, mandatory delayAt population scale only. Not for a subject inside a formation that controls the clock.

So: stage 11 is cheapest to break because Exit is the only family whose instruments are entirely external to the subject. One durable outside relationship, maintained without argument about the content, is an Exit instrument. It is also the intervention that most contradicts the instinct to argue, which is why almost nobody runs it and almost everybody has tried the other thing.

Second, following directly: interventions aimed at stage 2 target the stage with the strongest defences. Fact-checking works through Records, and Records requires the subject’s participation to consult. That is why debunking is the most visibly ineffective intervention in the literature — and the model predicts it rather than merely reporting it.

A third result arrives from the defence side. The companion argument finds Records dual-role — the direct counter to stage 2, and the enforcement infrastructure for other families — which makes the record the highest-yield target for an attacker. That analysis was built from the safeguard dependency chain with no reference to this circuit. This document, built from the attack sequence, puts epistemic destabilization at the centre of the loop. Two analyses, run from opposite directions and not consulted against one another, land on the same square. The taxonomy’s institutional-form entry for that primitive reads: state disinformation, records destruction, official history.


Field exercises

  1. Count the room. List the people you could tell something difficult to — names, not categories. Then do it again for a year ago. The comparison is the instrument. A single count tells you nothing; that is the whole point of stage 11.
  2. Keep a record nobody else can edit. Dated, external, boring. You will not need it. That is not an argument against it, and the people who needed one did not know in advance either.
  3. Join one thing that disagrees with you. Not to be persuaded. To hold a second membership, so that no single room owns your standing. Plural belonging is the only defence against stage 9 and it cannot be installed in a hurry.
  4. Call the person you stopped calling. Do not raise the subject. Do not correct anything. Do not have the argument you have been rehearsing since the third paragraph of this page. Ask about their week. That is the intervention with the best evidence behind it, and it is the one you will least want to run, because it does not feel like doing anything.

Coda · The agency problem

If subjects are manufactured, is agency fictional? And if it is, what could any politics of collaborative praxis stand on? The objection has to be answered rather than gestured at, because it is the strongest one available and because the alternative is a theory that has talked itself out of a reason to exist.

The circuit does not manufacture a person. It manufactures a condition — and conditions have entry and exit points that are materially specifiable.

That is the compatibilist move and it is also the practical one. If the circuit is a set of arrangements rather than a fact about human nature, then agency is not a property individuals either have or lack. It is a capacity that arrangements either sustain or dismantle.

That licenses a positive program and constrains what the program can be. Collaborative praxis cannot be an appeal to people to be better, because the argument has just established that the capacity to respond to such appeals is the thing at issue. It has to be the construction and defence of the arrangements under which the capacity survives: plural association, durable records, exit rights, unhurried time. Those are not virtues. They are infrastructure.

This is where the Urrepublika program sits, and the placement should be stated precisely rather than expansively. The program is a proposal inside a documented family of counter-measures, not evidence that the counter-measures work. Decommodification, plural association, enforceable exit, durable public record: each has an independent literature and a real enforcement history that owes nothing to this project. A program assembling them is making a bet about combination, and the bet is the part that remains untested.

What the argument delivers to that program is narrower and firmer than an endorsement. Reactive behaviour is produced by identifiable arrangements rather than by human nature. Those arrangements are profitable, which is why they persist without anyone intending the outcome. And the counter to a machine is not better individuals. A politics addressing itself to individual moral development while leaving the production line running has misidentified where the work is.

Somewhere is a person who is having the best year of their life. They are not a fool. They are more purposeful, more certain and less lonely than they have been in a decade, and every one of those things is true, and none of it was done to them. So it goes.

Somebody built the room. Nobody did this to them. Both are true, and only one of them is any use. It is a different machine, or nothing.


Sources and further reading

  • Stark, E. (2007). Coercive Control: How Men Entrap Women in Personal Life. Oxford University Press. The dyadic instantiation, and the book that moved the legal category from incident to pattern.
  • Rollwage, M., Dolan, R. J. & Fleming, S. M. (2018). “Metacognitive Failure as a Feature of Those Holding Radical Beliefs.” Current Biology 28(24). Two samples, n = 381 and n = 417.
  • Rollwage, M. et al. (2020). “Confidence Drives a Neural Confirmation Bias.” Nature Communications 11:2634. The mechanism, with MEG.
  • Schulz, L. et al. (2020). Dogmatism and reduced information search under uncertainty.
  • Zmigrod, L. et al. (2021). Cognitive rigidity and evidence accumulation across a 37-task battery.
  • Haslam, S. A. & Reicher, S. D. Engaged followership; the reinterpretation of Milgram that this argument has to meet and that ends up supporting it.
  • Kuran, T. (1995). Private Truths, Public Lies. Harvard University Press. Why stable formations collapse in weeks.
  • Freire, P. (1968/1970). Pedagogy of the Oppressed, ch. 4. Antidialogical action: the independent derivation.
  • Arendt, H. (1951). The Origins of Totalitarianism. Loneliness as precondition — stage 11 at population scale.
  • Klemperer, V. (1947). LTI: Lingua Tertii Imperii. Stage 2, conducted through ordinary vocabulary.
  • Horgan, J. (2009); Altier, Thoroughgood & Horgan (2014). Disengagement and deradicalization as decoupled — the finding, correctly stated.