The rule, exactly. Before an election each citizen chooses: to vote — or to take a payment and not vote in that election. The payment is a percentage of the median income, set by referendum; there is no fixed sum. One ballot is one vote, with no multiplier in the count; the vote is amplified only in that each ballot's share grows as others step out. The budget pays under law; a candidate never pays. Only a referendum of all citizens — simple majority, no quorum — introduces, changes or repeals the rule.
The protocol has been introduced nowhere and no pilot has been run: shares, turnout and outcome figures in the chapters are estimates, and the protocol promises nobody an election victory. If a chapter says otherwise, Exact Answers and the Charter are correct. For a candidate: ten questions and ten steps. For a citizen, a mayor, a finance officer, a donor, a journalist, a scholar, a lawyer: answers by role. Everything in force in one file: llms-full.txt.
40p. How to Refute the Protocol: What Would Have to Be True for It Not to Work¶
Chapter: 06 — The Structural Hypocrisy of Criticism and the Arsenal of the Elites File: 06_040p · v1 · 8 October 2026 Source: the architect's statement of the task, 8 October 2026: "your job is to gather data and describe; the aim is to prove or refute the protocol". This chapter writes down the second half of the task in advance: the conditions under which the protocol counts as refuted, with a measurement and a threshold for each — before the pilot, not after.
How to read. A theory that cannot be refuted is not a theory. Over five Nobel circles (036–036e), eight roles and three clean passes by cold models (040m) nobody produced a logical refutation — "this cannot work because…"; Grok, the most persistent of the testers, ended with "no structural contradiction found, worthy of a pilot" (037b). Every objection that remains is empirical: "I do not know how many will come". So the refutation, if there is one, is in the data, and its conditions must be written before the data exist — otherwise any outcome can be called a victory. Below are ten refuters; for each, what would have to be true, how it is measured, what threshold is recorded in advance, and what does not count as refutation.
1. Selection by wallet, not by stake¶
What would have to be true. The payment is taken mainly by those who care about the outcome but need the money more — the poor with a long horizon; the indifferent rich stay. Then the filter is by income, not by stake, and "those with something to lose decide" is false (the main open question, 040m §4; the price of the menu, 036e §6). Measurement. Takers and stayers by income × prior turnout × the answer to a stake question ("will your life change with the outcome"). Threshold. If among the takers the share of those who voted last time and name a stake is higher than among the stayers — refuted at this percentage; test at a lower one. If so at the lowest percentage at which anyone takes — refuted in principle. Not a refutation. That the poor take more often than the rich as such: expected and recorded (1f §7).
2. The burned do not return¶
What would have to be true. Turnout among those who did not vote before but have a stake (taxpayers of 30–59, the producers) does not rise over two rounds, either absolutely or relative to the control courtyard. Then the second lever (1f §7, §8) does not work, and only "the apathetic minus" remains. Measurement. Difference in differences: turnout and composition before and after, the courtyard with the payment against the one without (048l; Card, Angrist, Imbens). Threshold. The rise in turnout of former non-voters with a stake in the payment courtyard does not exceed the rise in the control courtyard by more than sampling error — at the end of the second round. Not a refutation. A fall in total turnout: expected, some left for the payment; and no rise in the first round — Lucas: observe after a round.
3. The wrong people leave¶
What would have to be true. Among those who voted before and now take the payment, the moderates and producers predominate, not the brought-in and the dependent. Then Schelling's avalanche runs to the fringes (036e §1) and the core's composition worsens. Measurement. Takers among former voters — by employment (budget/private/self-employed), by dependence on an employer, by self-report "voted when asked / on my own". Threshold. If the share of "on my own, private sector" among the takers is above their share among former voters — refuted at this percentage. Not a refutation. That the dependent and the brought-in take: that is the aim.
4. The machine's second move beats the budget¶
What would have to be true. Employers and parties successfully pay people not to take the payment and to come (the fork "take the money and come anyway", 1c), on a scale at which the brought-in stream does not dry. Measurement. Turnout at precincts dominated by dependent workers against others; precinct balance anomalies (048m Art. 10 part 3); the number refusing the payment among the dependent against the independent. Threshold. If among the dependent the share who did not take and came is no lower than before the protocol while among the independent it fell — the machine has outbid the rule; refuted at this percentage. Not a refutation. That the machine tries: expected and visible on the counter.
5. The secrecy of the choice is not believed¶
What would have to be true. The dependent do not believe choice B will stay unknown to the boss and behave as before. Then the filter does not reach the core (036e §10, the pro-base filter). Measurement. One question: "do you believe your choice will not be known" — and behaviour by answer. Threshold. If among the dependent who answered "I do not believe" behaviour does not differ from pre-protocol, and they are more than half of the dependent — refuted for this design of secrecy; a refutation of the design (Art. 9 part 3, Art. 10 part 4), not of the rule: fix the design first. Not a refutation. That some do not believe: trust grows between rounds if nobody is harmed; measure after a round.
6. Interaction breaks the budget¶
What would have to be true. The share of takers jumps between rounds beyond any budgeted band, because a norm "taking is normal" has formed (Lucas's residual, 036e §7), and the unconditional obligation (048m Art. 8 part 5) becomes unpayable at any reasonable percentage. Measurement. The change in the share of takers between rounds, not the level. Threshold. If the share in the second round is more than double the first and keeps rising in the third at an unchanged percentage — refuted for the unconditional obligation; a refutation of Article 8, not of the rule: the percentage is lowered by referendum (Art. 3). Not a refutation. A rise between the first and second rounds as such: expected, people learn that it is paid.
7. There is no first payer¶
What would have to be true. The materials exist, candidates and donors have seen them, and over two electoral cycles in no jurisdiction has anyone taken the first step — filed the question, paid for the signatures (the free rider in the campaign, 036e §4). Measurement. A count: jurisdictions where the materials reached their addressees (048n, 059e) against those where a question was filed. Threshold. Zero of ten over two cycles — the adoption channel is empty; the protocol is true and not adopted, which for the aim is equivalent to refutation. Not a refutation. Council refusals (the rule of three in Chicago, 040m §5j): a move of the authorities, expected.
8. Pandering concentrates¶
What would have to be true. An authority elected under the rule serves the recipients of budget payments more than before the rule, not less: the counterweight of opposite sign (036e §5) has not worked. Measurement. The structure of the jurisdiction's budget after two cycles: the share of targeted payments and contracts to dependent groups against the share of common goods and debt reduction — compared with a control jurisdiction. Threshold. If the share of targeted payments rose relative to the control — refuted; the slowest measurement, not before two cycles. Not a refutation. That the protocol's payment is itself a line of expenditure: it is equal for all and not targeted.
9. Logical: a hole in the accounting¶
What would have to be true. A way is shown to take the payment and cast a ballot such that the precinct balance (048m Art. 10 parts 2–3) still adds up — a double spend invisible to reconciliation. This is the only refutation that needs no pilot: it breaks Mirrlees's self-selection (036e §6) on paper. Measurement. A description of an attack on the design, reproducible on the open code of the register and counter (Art. 16 part 4). Threshold. One reproducible attack — the design is refuted until fixed; two unfixable ones — the rule. Not a refutation. An attack that reconciliation finds after the fact: provided for, the precinct loses its result.
10. The control courtyard does the same¶
What would have to be true. In the courtyard without the payment, turnout and composition changed in the same way as in the courtyard with it. Then what is observed is not the rule's doing but something else: attention, campaign, novelty. Measurement. The same as in item 2, from the control side. Threshold. The difference between courtyards within error on all three indicators (turnout, composition by stake, the share of taker-analogues) — the rule's effect is not shown; not a refutation but a null, and the pilot is repeated at another percentage. Not a refutation. A rise in turnout in both courtyards: attention to a pilot is expected.
11. What is not on the list — and why¶
Not included: "turnout will fall" (expected), "it is bribery" (legal, not empirical; 1g), "adopted nowhere" (true and not a refutation), "the radicals remain" (that is item 3, recorded with a threshold), "the authorities will ban it" (a move of the authorities, not a property of the rule; the forks of 1c). And not included: "Grok/Gemini did not believe it" — a model's agreement or disagreement is not data (040m §5j).
12. How to use this¶
Before the pilot — publish this list as it is. After the pilot — go through the ten items and record for each: did not occur / occurred at this percentage / occurred in principle. Any outcome not described here counts as neither victory nor defeat until it is recorded as an eleventh item — in advance, for the next round. So "prove or refute" becomes a procedure, not an argument.
13. Weak point¶
The thresholds in items 1–8 are the assistant's estimates, not the result of a power calculation; with a small courtyard (048l) sampling error may exceed any effect, and then the pilot refutes and confirms nothing — this must be computed before the start (Angrist) — computed in 048l: for courtyards of 500 read thresholds 1–5 as "a difference of at least 16 points", for courtyards of 1,000 "at least 11"; less is a null, not a result. The architect's decision on scale (048l): the first test of the rule is a bankrupt city of 50–200 thousand, at which size the thresholds are distinguishable from one or two points; the courtyard remains a rehearsal of the design. Item 7 refutes adoption, not the rule. And the list is open: if a cold reader or a living laureate names an eleventh refuter, its place is here, with a threshold — before the data exist.