The Hands on the Frame
Fascomachy · Document Three
The answer is paperwork.
Evidence tier: Tested. Falsifiers stated, and at least one has been run against the work. Evidence Protocol · Theme Explorer
Bottom line up front
Do not begin by deciding whether the person influencing you is good. Build conditions in which a bad actor cannot quietly control the frame, clock, record, exit, or cost of refusal.
The first two documents described manipulation and the circuit that can turn separate techniques into a self-repairing system. This document asks the practical question they left open: what makes influence legitimate?
There is no reliable test of intention. A teacher and a recruiter may use the same method, under the same rules, while claiming the same concern for your future. The safer answer is structural: change the room so an interested party cannot make the decision irreversible, unauditable, hurried, isolated, or too expensive to contest.
On this page
- Bottom line up front
- What did not happen to you
- 1 · The question the taxonomy refused
- 2 · Eleven safeguard families
- 3 · Remove the trusted vanguard
- 4 · The perturbation test fails
- 5 · No metric alone
- 6 · What broke under attack
- 7 · Falsification
- Good-Evening
- 8 · What follows
- Field exercises
- The last thing
- Sources and further reading
What did not happen to you
No official had to decide that you were manipulated. Nobody had to certify your vulnerability. A cooling-off period gave you morning. A funding register made borrowed authority checkable. An appeal created voice. Portable income made exit possible. The safeguards worked on the arrangement, not on your character.
1 · The question the taxonomy refused
The taxonomy catalogued 13 mechanisms but did not supply a moral sorting test. That refusal was correct. Intent cannot carry the distinction because every manipulator can claim benevolence and every legitimate educator still shapes attention.
Instead, ask which protections operate around the interaction. A cooling-off period defeats time pressure without harming an honest teacher. A funding register exposes hidden interests without silencing expertise. The identity of the influencer matters less because fewer consequences depend on trusting it.
2 · Eleven safeguard families
The 11 families group legal and institutional tools that already appear in statutes, professional rules, platform regulation, and parliamentary practice. Eight improve the conditions under which people judge. Three remove the capacity, technique, or transaction that carries the pressure.
| Family | What it changes | Typical instruments |
|---|---|---|
| Time | Delay between pressure and consequence. | Cooling-off periods; minimum reading time. |
| Provenance | Make the source, funding, and interest checkable. | Funding, lobbying, and gift registers. |
| Exit | Make leaving materially possible. | Portable benefits; independent income; cancellation rights. |
| Fixed Terms | Set the frame by rule, not by the interested party. | Pre-registration; fixed questions; standard contracts. |
| Records | Keep the past from being rewritten by the present. | Durable logs; version histories; public records. |
| Separation | Split roles that become dangerous when combined. | Recusal; non-voting chairs; blind review. |
| Voice | Let people contest a system without leaving it. | Appeal, representation, dissent, and review. |
| Symmetry | Limit the effort required to exercise a right. | One-click cancellation; pre-filled forms; tell-us-once rules. |
| Prohibition | Ban a technique or input outright. | Limits on dark patterns, intimidation, or specified uses. |
| Capability | Reduce an actor’s power to run a technique at scale. | Structural separation; access limits; divestiture. |
| Decommodification | Remove the sale on which the manipulation depends. | Public provision; commons; universal services. |
Confidence differs by family. Time, Provenance, Exit, Fixed Terms, Prohibition, Capability, and Decommodification have established examples. Records overlaps with Provenance. Separation’s internal boundaries remain contested. Voice was recovered through external review. Symmetry has outside support but did not reproduce in the internal grouping method.
Primitive mapping
| # | Primitive | Covering families |
|---|---|---|
| 1 | Frame control | Fixed Terms · Separation · Voice |
| 2 | Epistemic destabilization | Records |
| 3 | Affect induction | Time · Prohibition |
| 4 | Obligation manufacture | Provenance · Prohibition |
| 5 | Commitment capture | Time · Fixed Terms · Separation |
| 6 | Consensus fabrication | Provenance · Capability |
| 7 | Authority borrowing | Provenance · Separation |
| 8 | Time compression | Time |
| 9 | Belonging pressure | Exit · Voice |
| 10 | Attachment exploitation | Exit · Symmetry |
| 11 | Isolation and dependency | Exit |
| 12 | Choice architecture | Fixed Terms |
| 13 | Staged demonstration | Records · Provenance |
For the concrete surface — the roughly ninety named tactics people actually recognize, sorted by arena and mapped back to the primitive each one instantiates — see The Grammar of Manipulation, Part III — Tier 2: The Named Tactics.
Four primitives have only one line of defence: epistemic destabilization depends on Records; time compression on Time; isolation on Exit; and choice architecture on Fixed Terms. These are the thin places. If single-covered primitives prove no easier to exploit than multiply covered ones, the claim that safeguards combine multiplicatively is wrong.
The capture circuit
The circuit enters through isolation, destabilizes the record, supplies belonging, and accelerates judgment through affect. Exit and Voice help prevent the sequence from closing into a loop. Records protects the archive. Time protects the clock. If Records and Time fail without capture following, or Exit and Voice remain strong while capture closes anyway, this account needs revision.
Reversibility is an outcome
A right to reverse a decision is useful only while it remains usable. Time keeps the window open. Exit prevents the relationship from controlling survival. Symmetry prevents cancellation from costing five calls when signup took one click. Remove any one and reversibility may remain legally present while becoming practically decorative.
Records supports other safeguards
Records directly counters destabilization and also enables Provenance and Symmetry. A funding disclosure or burden rule that leaves no durable trace cannot be enforced. This makes the record an efficient first target: damage it once and several safeguards weaken together.
What works unattended
Some safeguards execute themselves: sunset clauses, secret ballots, transfer delays, fixed questions, and cooling-off periods. Others exist only when enforced. An unaudited register is not a weaker safeguard. It is a document describing one. Provenance and Records are especially vulnerable to this conversion.
3 · Remove the trusted vanguard
Freire, Habermas, and Mouffe each identify something worth protecting in dialogue: agency, contestation, and freedom from domination. The structural problem appears when a good facilitator remains the guarantee. A system that works only when its operator is decent has not solved the operator problem.
Split the role. The person who convenes, sequences, or summarizes should not originate positions. Professional facilitation, non-voting chairs, blind review, adversarial collaboration, recusal, and pre-registration use versions of this rule. Each can still be gamed, especially through speaker order, admissibility, timing, and summary. The point is not purity. It is a checkable limit where trust used to stand.
4 · The perturbation test fails
You cannot detect manipulation by challenging a belief and watching whether it moves. Strong expertise can remain stable under attack. Ordinary reasoning also shifts supporting reasons to fit conclusions. Resistance may increase certainty, especially among people reasoning carefully.
A composite assessment may examine metacognitive sensitivity, the direction and cost of changing reasons, and group-level links between reasons and preferences. None can diagnose a person. Base rates and thresholds are unknown, and the most dangerous error would be classifying a well-grounded dissenter as captured.
5 · No metric alone
Goodhart’s and Campbell’s laws apply here. Any legitimacy score can become a new instrument of control once people optimize against it. Pair quantitative indicators with qualitative evidence of human effects, and let neither justify coercive change alone.
An arrangement can score well across all 11 families and remain illegitimate. The families describe necessary supports, not a certificate. They will eventually appear in a maturity model with traffic lights. The traffic lights should not be permitted to decide.
This account also over-predicts manipulation in contemporary commercial life because many commercial environments lack cooling-off time, visible interests, effort parity, and meaningful exit. That is a conclusion. Severity still varies. The useful question is which families are missing and who had the authority to build them.
6 · What broke under attack
Seven attacks were run. Four changed the argument. The original three safeguards could not cover several structural counters, and tempo defeated all three. The neat one-family mapping collapsed into many-to-many coverage. Staged demonstration appeared as a missing primitive. An external review recovered Voice and then exposed Fixed Terms and Prohibition as missing families.
Symmetry did not reproduce in two cold grouping runs. External evidence still supports regulating administrative burden as an object separate from eligibility or benefit terms. The page therefore keeps Symmetry with lower confidence and a live falsifier rather than selecting the more flattering result.
7 · Falsification
The account fails if a successful manipulation case cannot be explained by any family’s absence; if single-covered primitives are no more exposed than multiply covered ones; if Records collapses completely into Provenance; if structural remedies perform no better than behavioural ones across a credible sample; if decommodified provision reproduces the same commercial primitives administratively; or if enforced facilitator neutrality makes no measurable difference.
Symmetry should be regrouped with enough examples—effort parity, pre-population, tell-us-once rules, and simplified submission—to match the density of other families. If it still fails to appear under a materially different method or human review, the count should fall to 10.
Good-Evening
Fiction. The plant, the incident and everyone in it are invented. It illustrates the argument; it is not one of the cases this page tests.
Statement recorded after the frame failed.
They called me Good-Evening because I arrived after every decision and before every consequence. This was not my title. My title was Safeguard Liaison, but nobody outside the building could remember it, and nobody inside used names unless an action had to be assigned.
The frame stood around the machine. It was painted yellow and bolted to the floor. Visitors were told never to place their hands beyond it. Operators were told the same, except when production required otherwise. Production required otherwise often enough that the warning became a description of who would be blamed.
My work was to verify the safeguards. The emergency stop existed. The appeal channel existed. The incident record existed. The independent reviewer existed two provinces away under a contract renewed by the office being reviewed. Each item could be found. This was called assurance.
Mara worked the late shift. The machine began drawing material sideways, which it had done twice before, though Engineering had not classified either event as a recurrence.
She reached for the stop. A locked guard covered it because an earlier audit had found that workers could activate it accidentally. The shift supervisor held the key. He had gone home.
She called me. I said good evening.
She said the machine had her sleeve. The procedure allowed the guard to be forced in an emergency, but only on a supervisor’s instruction. I could hear the frame moving on its bolts and Mara saying she could not pull free. The supervisor did not answer. I opened an incident record because that was within my authority.
By then two workers had arrived. One held Mara. The other struck the guard with a wrench. The guard broke. They stopped the machine.
The safeguard system later classified the wrench as an unauthorized intervention and the broken guard as damage to protective equipment. Mara lost two fingers. The report said the injury was limited because the emergency stop functioned when activated. It said workers had bypassed a guard. It said liaison support had been available throughout the incident. All three statements were true.
At the review I argued that a safeguard no one could use under pressure was not a safeguard. Legal asked whether the stop had failed. I said no. Operations asked whether the appeal channel had failed. I said no. Compliance asked whether any required record was missing. I said no.
The chair concluded that the system had operated as designed and recommended refresher training on keeping hands outside the frame. Mara attended by video because she had not yet returned to work.
When my turn came, she raised what remained of her hand. The chair did not see her. The participant window was behind the minutes.
I said good evening.
8 · What follows
Start with the primitive creating the harm. Identify the safeguard families that cover it. Prefer defaults and structural constraints to vigilance. Add enforcement where the instrument requires it. Record the cost, test exit, and publish the failure path.
Plain language belongs at the point where a person exercises a right. It lowers the labour required to cancel, appeal, correct, or leave. Plain wording elsewhere may help; plain wording at the point of action changes whether the safeguard can be used.
The positive program is institutional: preserve time, exit, records, voice, symmetry, and role separation; prohibit techniques where conditioning cannot contain them; remove capabilities that make abuse scalable; decommodify necessities where the sale itself carries the pressure.
Field exercises
Map a decision. Name the frame, clock, record, source, exit, appeal, and cost of refusal. Break the trusted-person test. Ask whether the safeguard survives a competent bad actor. Run Tuesday against Wednesday. Identify what lets a decision made under pressure be inspected after the pressure has passed.
The last thing
The hand on the message matters. The hand on the frame matters more. Build the door before deciding whether the person beside it means well.
Sources and further reading
- Hirschman, A. O. (1970). Exit, Voice, and Loyalty. Harvard University Press. Supplied the family that had gone missing.
- Freire, P. (1968/1970). Pedagogy of the Oppressed, ch. 4. The structural mirror, and the vanguard problem.
- Mouffe, C. (1999). “Deliberative Democracy or Agonistic Pluralism?” Conflict and passion as constitutive rather than contaminant.
- Habermas, J. Communicative versus strategic action — taken without the ideal-speech idealization.
- Herd, P. & Moynihan, D. (2018). Administrative Burden: Policymaking by Other Means. Russell Sage. The Symmetry evidence base.
- Rollwage, M., Dolan, R. J. & Fleming, S. M. (2018). “Metacognitive Failure as a Feature of Those Holding Radical Beliefs.” Current Biology 28(24).
- Niemeyer, S. et al. (2023). “How Deliberation Happens: Enabling Deliberative Reason.” American Political Science Review. The Deliberative Reason Index.
- Goodhart, C. (1975); Campbell, D. (1979). Why no legitimacy metric may be read alone.
- Fels, A. On structural versus behavioural remedies in competition law — the fifty-year-old version of this document’s type distinction.
- European Commission. Merger Remedies Study — roughly a third wholly or partly ineffective. The caveat that belongs in the same breath.