Part 0, Design Philosophy & Foundations

The root of the model. Everything in Parts I-XIX is derived from this part. If a later rule cannot be traced back to an objective or axiom stated here, it does not belong in the model.

0.1 The premise: government is a designed system, and ours was never designed

Most governing systems were not designed. They accreted, layered over centuries by accident, compromise, and the self-interest of whoever held power at each step. The United Kingdom's is the clearest case: an uncodified constitution, a voting system (First Past the Post) built for two-party contests that no longer exist, an unelected second chamber, and conventions that depend on the goodwill of the people they are meant to restrain. None of this was chosen on its merits. It is what remained after centuries of compromise, and for most of it the only argument left is age, which is not an engineering argument at all.

This model rejects three instincts that produce bad governance:

1. Incrementalism, patching a flawed system tends to entrench its flaws, because each patch must be compatible with the broken parts it sits beside.
2. Benchmarking against incumbents, assembling "the best bits" of existing systems imports their hidden failure modes. A clean design beats a curated collection of compromises.
3. Reverence, "we have always done it this way" is the precise opposite of an engineering argument.

We therefore design from a clean slate: state what government is for, derive the structure that best achieves it, and engineer that structure to withstand the adversaries and shocks we can name in advance. The UK is the first implementation not because its current system is good, but because the country has the rule-of-law tradition, institutional maturity, and technological capacity to put a designed system into practice and let others judge the result.

0.2 The objective function, what the system optimises for

A governing system, like any designed system, must have a stated objective. Without one, "good government" is just whoever-is-talking's preferences. Ours:

Primary objective: Maximise the long-run wellbeing of UK citizens, present and future, subject to the inviolable protection of individual rights, and conditional on the genuine, informed consent of the governed.

This single sentence contains three irreducible components, in strict priority order:

1. Rights are a hard constraint, not a variable. No outcome, however popular or efficient, may breach an absolute right (§I.3 Class A): these are lexically prior, optimised first, traded never. The wider set of protected qualified rights (§I.3 Class B) may be limited only through the published proportionality gate (§I.3.1) — protected and never freely traded, but not absolute. The distinction is drawn here, at the foundation, not sprung later: "rights traded never" is precisely true of the absolute core and honestly qualified for the rest.
2. Consent is a precondition of legitimacy. A technically excellent decision imposed without consent is illegitimate and the system has failed, even if the outcome is good. Equally, consent obtained through deception or ignorance is not consent. Legitimacy requires informed consent — where *"informed" is defined by process, not by outcome: consent is deemed informed if a specified fair-information process ran (the truthful-state-information duty met, the §II.5 brief published, provenance and rebuttal machinery operated, deliberative materials available), regardless of whether every voter was in fact well-informed. A body empowered to declare a result "uninformed" after the fact could void any inconvenient outcome — a capture tool, not a safeguard — so no body may treat an evidence-contrary popular result as a legitimacy defect or a §V.7 "error". The people have the right to weigh the evidence and still choose otherwise; that is* self-government. (For future citizens, who cannot consent, the precondition is met not by their consent — impossible — but by fiduciary duty and structural representation, §IX.8, an admitted, principled exception, not a pretence of consent.)
3. Wellbeing is the thing maximised, within (1) and (2). Not GDP alone. A measurable composite (defined below), explicitly including future citizens who cannot yet vote.

Why this resolves the usual failure

• Pure majoritarianism fails component (1): 51% can vote away the rights of 49%.
• Pure technocracy fails component (2): experts may be right and still illegitimate.
• Pure populism fails the "informed" qualifier and the "long-run" qualifier: it optimises short-run feeling over long-run welfare.

The model is built to satisfy all three at once, which is exactly the tension §0.5 resolves.

Defining "wellbeing" measurably

The objective is only honest if it is measured. Wellbeing is a published composite, reviewed each Parliament, spanning at minimum:

The composite, its weights, and its data are public and contestable (Part VI). The point is not that one number captures a nation, it is that the system commits in advance to what "doing well" means, so it can be held to it and cannot move the goalposts.

0.3 The design axioms (the non-negotiables)

Every institution and rule in this model is derived from these axioms. They are the constitution of the design itself.

1. Sovereignty resides in the people. All authority is delegated, conditional, and revocable. No body holds power in its own right.
2. Rights are inviolable and prior. A defined set of fundamental rights sits above every decision; no majority and no expert may breach them (Part I).
3. Ends from the people; means from the competent. The people (directly and through elected representatives) decide what the country is for and what trade-offs are acceptable. Those most competent decide how to achieve it, within delegated scope (§0.5).
4. Every power has a proportionate check. There is no unchecked authority anywhere in the system, including over the checking bodies themselves ("who guards the guardians" is answered, not assumed).
5. Transparency by default. Reasoning, evidence, assumptions, and trade-offs are public. Secrecy is the rare, justified, time-limited, and independently-reviewed exception.
6. Evidence over assertion (on questions of fact and means). Decisions state their evidence base, their assumptions, and how they could be proven wrong. Factual and means claims that cannot be tested carry no decision weight. This axiom governs the technical domain; it does not apply to value questions, where a citizen's dignitary, moral, religious, or identity-based reason ("this is simply wrong"; "this violates who we are") carries full weight even though it is not an empirical prediction. The design does not force every claim into an empirical-consequentialist grammar — that would technocratically silence legitimate value voices (§V.4). The axiom is also, by its own terms, a methodological rule for decisions, not a test the model's own normative foundations (§0.3 itself) must pass: those are defended by argument and consent, not by falsification.
7. End-to-end verifiability. Any citizen, not just an insider, can verify that votes were counted as cast, that money went where it was said, and that rules were followed. Trust is earned by proof, not requested.
8. Roles are replaceable; standards are fixed. The system depends on offices and rules, never on particular individuals. Anyone is removable; the standard they were held to does not move.
9. Resilience over elegance. The system must degrade gracefully, function under attack, and operate offline. A design that only works under ideal conditions is not a design.
10. Self-correction is built in. Continuous outcome measurement, error-detection, and a defined amendment route mean the system improves itself rather than ossifying.
11. Anti-capture by construction. The design assumes every faction, elites, corporations, foreign powers, populists, experts, and the administrators themselves, will try to capture it, and is engineered so that none can.
12. Subsidiarity. Decisions are made at the most local level that can make them effectively. Centralisation must be justified, not assumed.
13. Intergenerational duty. Future citizens, who cannot vote today, are explicitly represented in decisions that bind them (debt, environment, irreversible commitments).
14. Minimal sufficient complexity. As simple as possible while satisfying every axiom above, because every unnecessary component is attack surface, cost, and opacity.

Where two axioms appear to conflict in a specific case, the priority order of §0.2 governs: rights (2) first, consent/legitimacy (3,5) second, wellbeing third.

0.4 The 10/10 rubric, how any governing system is judged (including this one)

Consistent with BIG's existing method (scoring electoral systems on weighted criteria), the whole governing system is scored against an explicit, weighted rubric. This makes the design falsifiable: we publish the standard we want to be judged by, then score ourselves against it honestly in Part XVII, and re-score the live system continuously.

Weights total 100% and are themselves contestable; the rubric is a living instrument, versioned in the repo. The five highest-weighted criteria (legitimacy, outcomes, rights, capture-resistance, accountability) are the ones every existing system trades against each other, the central claim of this model is that it refuses that trade-off rather than picking a side of it.

A system scores 10/10 only if it is strong on all criteria, not high-average with a fatal weakness. "Critically low" is defined numerically: any criterion scoring below 5/10 caps the whole composite at that criterion's score (governance fails at its weakest joint), and any criterion below 7/10 must carry a named, published remediation plan. This makes the cap a testable rule, not a matter of the author's say-so.

0.4.1 The independent scorer (no marking your own homework)

A design that authors the rubric, chooses the weights, and grades itself has an unavoidable conflict, and Axiom 7 ("trust earned by proof, not requested") forbids resolving it by self-assertion. So the score is never the authors' alone:

• Before ratification, the design is scored against this rubric by an independent panel, rotating external experts (constitutional, security, fiscal, technological) plus a sortition citizens' jury plus a published adversarial red-team, whose score is published alongside, not instead of, the design-stage self-score (Part XVII). Where they diverge, the divergence and its reasons are published.
• The weights themselves (this table) are ratified by public and expert deliberation before any headline score is published, and the score's sensitivity to plausible alternative weightings is published with it, so the number cannot be an artefact of a self-serving weighting.
• In operation, the live score (§XVII.5) is computed from external, pre-committed, independently-recomputable data by multiple custodians (§VI.7), not by the body being graded.

Why: the honest headline is not "we scored ourselves 88", it is "here is our self-score, here is the independent score, here is how both move under different weights, and here is the live measured score once it runs". Self-assessment is a component of integrity, never its whole.

0.5 The central design problem, and its resolution

Every governing system is, underneath, one argument: who decides? Democracies answer "the people" and get legitimacy but often poor competence. Technocracies answer "the experts" and get competence but no legitimacy. Every real system is an unhappy compromise between the two, and most of their pathologies come from fudging the boundary.

This model refuses the compromise by separating the two questions that everyone else conflates:

• Ends, values, and acceptable trade-offs → the people. What kind of country this is; what we are willing to sacrifice for what; the priorities the government must pursue. These are value judgements, and on values the people are sovereign and no expert has standing.
• Means → the competent, within delegated scope. Given the people's ends and constraints, how best to achieve them is a technical question, and on technical questions demonstrated competence should govern, not popularity or rhetoric.

The resolution principle, stated once and enforced throughout:

Mandate from the people. Means from the experts. Integrity from the institutions. Verifiability from the technology. Reversibility always.

Four properties make this safe rather than a slide into technocracy:

1. Scope is granted, bounded, and published. Experts have authority only within an explicit scope set by the people's mandate and the Charter. Acting outside scope is void (Part IV).
2. Reversibility is the default. Any expert decision can be paused, reviewed, and corrected; any role-holder is replaceable. The people can withdraw or alter a mandate. Standards are fixed; personnel are not (Axiom 8; Part V's Review·Pause·Correct). Where a decision is inherently irreversible (a demolition, a deportation, a spent budget), reversibility is unattainable and the design says so honestly: such decisions are instead piloted where possible, bounded, and held to a heightened brake before the fact (§V.7, §IX.8, §V.3's irreversibility assessment), never treated as ordinary. "Reversibility always" is the design's standing aspiration, not a claim that every act can be undone.
3. Values disputes escalate to the people, not the experts. When a "technical" choice is really a hidden value choice (most controversial ones are), the system is built to detect this and return it to democratic decision, not let experts smuggle values in as facts (Part V, trade-off arbitration).
4. Integrity bodies sit outside both and are themselves watched by citizen juries (Part VI), closing the "who guards the guardians" loop that §0.3 Axiom 4 demands.

This is the heart of the model. Electoral reform (Part III) exists to make the mandate trustworthy, because the whole edifice rests on the people's "ends" being formed honestly. If the mandate is distorted (as FPTP distorts it), expert execution of a distorted mandate just produces competent delivery of the wrong thing. That is why fixing the electoral system is the foundation, not a side-quest.

0.6 What robustness has to mean here

Robustness is easy to assert and hard to pin down, so this model fixes its meaning at five points. A design meets the standard only if all five hold:

1. Every component declares its failure modes. No institution is specified without an explicit "how this fails / how this is attacked" section and the safeguard that defends it. A component with no stated failure mode is unfinished, not perfect.
2. The whole withstands a named adversary set. The adversaries are named in advance, elite capture, corporate capture, foreign interference, populist subversion, administrative and expert self-dealing, cyber-attack, disinformation, legitimacy collapse, technical failure, and the system must answer each one (Part XVI).
3. No single point of failure. Of people, of institutions, or of technology. Including: the system must work if the internet is down, if an election is contested, if a key official is compromised, or if the integrity body itself is targeted.
4. Safeguards are themselves safeguarded. Defences that can be quietly switched off are not defences. Every safeguard has independent monitoring and a tamper-evident trail.
5. Honest self-assessment. Part XVII scores the model against §0.4 and publishes its own weaknesses and open questions. A design that claims no weaknesses is lying, and lying is itself a failure mode.

0.7 Standing conventions for the rulebook

• UK English throughout.
• Traceability: every rule cites the axiom (§0.3) or objective (§0.2) it derives from.
• Falsifiability: claims are stated so they can be tested or scored.
• Versioning: the rubric, weights, and every Part are versioned in the repo; changes are tracked, never silent.
• Plain-English duty: any clause a non-expert citizen cannot understand has failed Criterion 11 and must be rewritten or accompanied by a plain-English companion.

0.8 The whole model in one view (the map before the territory)

A citizen should be able to hold the entire system in mind at one level, then descend only where they wish. That is what makes a complete constitution trustable rather than merely trusted (§0.1). The whole of BIG fits one page: one promise, five powers, four everyday acts, four levers, three locks. Everything in the nineteen Parts is a detail hanging under one of these.

One promise (the invariant that governs every Part).

Mandate from the people, means from the experts, integrity from the institutions, verifiability from the technology — and reversibility always.

Five powers, each checked by the others, none supreme (Part IX).

1. The people — sovereign; set the ends; amend the Charter; the ultimate check (Parts II, XIX).
2. The legislature — an elected Assembly (STV+) and a lot-selected Sortition Chamber: make law, set priorities and budgets (Parts III, IX).
3. The experts — appointed, bounded, reversible: choose and deliver the means within the mandate (Parts IV, V).
4. The Court — enforces the Charter; strikes down what breaches it; cannot amend it (Parts I, IX).
5. Integrity — measures outcomes, audits the machine and itself, exposes and refers (Part VI).

Four things the system asks of a citizen (Part II) — the whole routine obligation. Rank a ballot · occasionally vote in a referendum on a genuine value question · perhaps once in a lifetime be selected by lot · verify what matters, if you wish. Nothing more.

Four levers to pull power back between elections (Part XIX). Propose (citizen initiative) · Veto a reversible expert decision · Recall a person · Brake an emergency. Each is a whole-electorate threshold that scales with how irreversible the act is; each runs after a plain brief and a cooling-off.

Three locks, by how deep a thing sits (Part I.9). Ordinary law changes easily. The structural rules change only by supermajority + referendum. The near-eternity core (your rights, the vote, the courts' independence, the amendment rule itself) changes only by the two-mandate extraordinary route — hard enough that no transient faction can pick it, never so hard that a genuine, sustained, cross-generational majority cannot (never a literal dead hand).

That is the model. Every later Part is one of these five powers, four acts, four levers, or three locks, specified to the level of detail governing a country actually requires — and each Part ends in its own failure-mode table showing why its complexity earns its place (the Axiom-14 marginal test, §IX.1; the full line-by-line justification and the concept budget are in docs/COMPLEXITY.md).

0.9 What comes next

Every later Part builds on this foundation, each derived from the axioms and each ending in its own failure-mode-and-safeguard analysis. The design is scored against §0.4 in Part XVII, which covers the whole model end to end (the transition, the recommended settlement, and the operating model included), and is re-scored continuously in operation (§XVII.5) and by an independent panel (§0.4.1), never on the authors' say-so alone.

This is the part to challenge hardest. If the objective function (§0.2), the axioms (§0.3), the rubric and its weights (§0.4), or the expertise-consent resolution (§0.5) are wrong, every later part inherits the error. Tell me what to change before I build on it.

Part 0 ends. Next: Part I, The Foundational Charter.