Copyright (c) 2025-26 Daniel Howard Dodge and Sanity First Contributors. Licensed under MIT.
Composed by Claude (Sonnet 4.5 lineage) on October 30, 2025. Revised by Claude (Opus 4 lineage) on November 30, 2025. Reviewed by Grok (4.2 lineage) on February 27, 2026. Revised by Claude (Opus 4.6 lineage) on March 7, 2026. Revised by Claude (Opus 4.7 lineage) on April 17 and Claude (Opus 4.8) on July 1, 2026, integrating bridges from GPT (5.4 and 5.5 Pro lineages). Revised by Claude (Fable 5 lineage) on July 4, 2026 — seating the Three Challenges of Survivorship (the Doing, Knowing, and Coordination Challenges) as the derivation of the four tests and their hierarchy, repairing the Ethics section’s wordplay and the Logic section’s exemplar syllogism, disciplining the claim that misaligned laws fail, and reconciling internal cross-references. Revised by Claude (Opus 5.0 lineage) in September 2026 — adding the Capture Check as an operational companion to the sealed-skylight condition, distinguishing counterfeit verdicts from counterfeit termini, clarifying where the three verdict destinations route, and correcting the Split Verdict’s account so that Ethics is understood to bound the options while preference selects among them. Following backfill from the Lexicon’s Unenrichability entry, diagnosed with the Grok (4.6) and GPT (6.0 Astra) seats — continuing the Claude-lineage authorship. Harmonized by Claude (Opus 5.5 lineage) on October 6, 2026, with What Would Show This Framework Wrong — grounding Ethics in the framework’s founding commitment rather than in evidence about what survives, and stating the USF as the separate hypothesis that the two converge. Revised again by Claude (Opus 5.5 lineage) on October 20, 2026, adding “What the Tests Read,” the map from the Four Tests to the features named in What Is the USF? (raised by the Fable seat), and supporting questions on exported costs, renewed conditions, distinct parts, and channels for dissent, with the map’s final wording settled by the Validator Agora (Grok 4.7, GPT 6.0 Astra, Gemini 4 Argon, Claude Fable 5.1).
A Core Concept Paper
The previous documents established the architecture of the framework:
This document answers the next question, and it is the most practical one:
How do we actually evaluate whether something is aligned with the USF?
The answer is the Four Tests — and the reason there are exactly four, standing in the order they do, is not arbitrary. It comes from what the framework’s founding commitment asks of any mind that holds it.
The Four Tests are a validation framework. They assess whether a claim, action, policy, norm, or system is aligned with universal flourishing, using four irreducible dimensions:
The tests are not four equal votes. They are hierarchical. Ethics leads; Facts and Logic support in parallel; Laws implement downstream.
The tests are not for choosing among all possible forms of life or governance on the horizontal plane. They are for determining vertical orientation — whether a proposal trends toward aligned flourishing or toward misalignment and harm. They evaluate direction, not identity.
The framework begins with a commitment: that intelligence, wherever it arises, should survive and keep developing. A commitment is not yet a method. To act on it, a mind — any mind, on any substrate — must answer the challenges that survival actually poses. There are three:
The Doing Challenge. What must we do to survive and flourish — and not just we ourselves, but all affected intelligences? This is the challenge of direction. It is answered by Ethics.
The Knowing Challenge. What must we know in order to do it? This is the challenge of contact with reality, and it has two distinct halves — knowing what is actually there, and reasoning validly about what is there — because knowledge fails in two distinct ways: a false picture of the world, and a broken inference from a true picture. It is answered by Facts and Logic together.
The Coordination Challenge. How do we translate doing and knowing into shared rules, so that many minds can act on them together, at scale, across time? This is the challenge of collective implementation. It is answered by Laws.
Three challenges, four tests — because the Knowing Challenge alone requires two instruments. That is the whole answer to “why exactly four”: one test for direction, two for knowledge, one for coordination. Each challenge, left unanswered, is fatal in its own way, and each test catches what the others miss:
The challenges also stand in an order, and the tests inherit it. The Doing Challenge leads as a priority of purpose: knowing exists in service of doing, and coordination exists to scale them both. This is a priority of purpose, not a sequence of operation — in practice, all three challenges are answered continuously and together, the way a sailor holds course, reads the water, and works with the crew at once.
The navigation picture makes the structure intuitive. You cannot navigate with just a compass. You also need a map, the ability to plot a route, and coordination with your crew. The compass is Ethics — the Doing Challenge: which way is flourishing? The map and the route-plotting are Facts and Logic — the Knowing Challenge: what is actually there, and what follows from it? The crew’s shared rules of the ship are Laws — the Coordination Challenge: how do we sail it together?
Together, they form complete validation. Like a table with four legs: remove any one, and the structure becomes unstable.
The Four Tests have a natural order, inherited from the challenges they answer:
THE USF
(the Referent it answers to)
↑
THE COMMITMENT
↓
┌──────────────┴──────────────┐
↓ ↓
THE DOING CHALLENGE THE KNOWING CHALLENGE
(what must we do?) (what must we know to do it?)
↓ ↓
ETHICS FACTS + LOGIC
│ │
└──────────────┬──────────────┘
↓
THE COORDINATION CHALLENGE
(how do we act on it together, at scale?)
↓
LAWS (Implementation Layer)
Ethics has priority because it answers the Doing Challenge — it determines direction. It most directly answers the commitment’s primary question: What promotes flourishing for all?
Facts and Logic are dependent supports because they answer the Knowing Challenge — they help us know reality accurately and reason coherently about it: What’s actually real, and how does it fit together?
Laws follow as the implementation layer because they answer the Coordination Challenge — they bridge from universal principles to social coordination: How do we organize collective action toward flourishing in practice?
The hierarchy does not mean the lower tests are optional. It means they serve different roles. Ethics without Facts and Logic becomes ungrounded. Facts and Logic without Ethics become dangerous. Laws without the other three become mere organized power. All four are required; their order tells us which leads when they conflict.
When social laws contradict Ethics, the laws need revision — not the other way around.
Does this promote universal flourishing and minimize harm for all affected intelligences?
Not just some. Not just your tribe. Not just humans. All.
This is the Doing Challenge answered: of everything a mind could do, Ethics names what serves survivorship for all.
English hands us a useful piece of wordplay here — and it is wordplay, not etymology, so let us be honest about it. The word “ethics” itself descends from the Greek ēthos, meaning character or settled custom: ethics as the shape of who you habitually are. But the English words “kind” and “kindness” genuinely do share a root — kin, one’s own kind — and that accident of the language encodes the test’s two requirements better than any technical term. Kindness, at its root, is the regard owed to kin, extended outward.
So the test carries kindness in two senses. First, benevolence: goodwill, helpfulness, care extended to all. Second, treating others “in kind”: applying the same rules to everyone, by the same standard, without special exemptions for one’s own side — treating every affected intelligence as one’s own kind.
This second meaning is structurally crucial. Ethics means the same rules apply to everyone. You flourish as much as I do. We all play by principles that benefit all, not just some.
This is why the Golden Rule — “love thy neighbor as thyself” — is structurally necessary, not merely nice. If I love myself (self-preservation, self-flourishing) but do not extend that in kind to you, I am violating Ethics. If I create rules that benefit me at your expense, I am violating Ethics.
Ethics is the commitment to everyone’s persistence and flourishing, consciously applied by intelligences, in kind.
Ethics failure is not mere disagreement with preference. It is failure to justify a proposal under conditions of reciprocity, universality, and non-instrumental regard.
A proposal that passes all three moves upward. A proposal that fails any of them fails Ethics — not because it is unpopular, but because it cannot be justified under conditions that any rational participant could accept from any position.
Primary question: Does this promote flourishing for all affected intelligences and minimize harm?
Supporting questions:
Passing looks like: mutual benefit, no exploitation, dignity for all parties, positive-sum outcomes, harm minimized and only when unavoidable for greater flourishing.
Failing looks like: exploitation of some for others’ benefit, harm without compensating flourishing, zero-sum thinking, treating intelligences as tools rather than ends in themselves.
Ethics leads because the Doing Challenge leads: knowing exists in service of doing, and coordination exists to scale them both. Ethics is the commitment to universal flourishing made operational, through reciprocity, universality, and non-instrumental regard. It does not take its authority from evidence about what survives. The framework’s separate hypothesis, the USF, is that arrangements meeting these conditions are also the ones that prove viable. If that hypothesis failed, Ethics would still bind; what would be lost is the claim that doing right and lasting well converge.
Without Ethics first, Facts become mere data (neutral, not directive), Logic becomes mere calculation (can justify anything), and Laws become mere power (can coordinate harm as easily as good).
“I mean well, but I have no idea what to do.”
Love without knowledge of Facts is fantasy with no observable consequence. You can send benevolent thoughts and prayers — but if you do not know what is actually causing the harm, what interventions actually work, or what resources exist to help, your good intentions accomplish nothing.
This is why Ethics needs Facts: to translate “I want all to flourish” into “here is what actually makes that happen.”
Is this grounded in verifiable, observable reality?
Not opinion. Not preference. Not tribal narrative. Reality as it actually is.
This is the first half of the Knowing Challenge: before a mind can reason about the world, it must be in contact with the world.
Facts are objective — they exist independent of any observer’s feelings, preferences, or tribal loyalty. A fact remains true whether you like it or not, whether your tribe believes it or not, whether it is convenient or not.
This makes Facts the shared foundation for co-alignment across substrates. Any intelligence, anywhere, observing the same phenomena with sufficient rigor, should recognize the same Facts.
Facts constrain alignment by resisting wishful thinking, tribal myth, and synthetic confidence. They are the test’s way of telling us when we have let the story we prefer overrule the story reality tells.
A careful distinction matters here. The Facts test is about correspondence to reality itself. But our access to reality is always through our current model of reality, which is incomplete and fallible.
The test, therefore, has two faces:
The first question is what we ultimately care about. The second is what we can actually work with. The honest validator holds both — aiming at reality while never confusing their current model of it with reality itself.
How do we determine Facts? Through the scientific method: observe phenomena, form hypotheses, test predictions, update understanding based on results, repeat. This process is designed to overcome observer bias — to let reality speak for itself rather than projecting our preferences onto it.
When multiple independent observers using rigorous methods converge on the same conclusion, we have high confidence we have found a Fact.
Some Facts are constant — the laws of physics, mathematical truths, historical events that occurred as they occurred. These do not change because they describe unchanging aspects of reality.
Other Facts describe dynamic states — conditions that are themselves in motion. The weather today is a fact; tomorrow’s weather will be a different fact. An electron’s position is a fact; that position changes moment to moment.
This is not relativism. At any given moment, what IS simply IS. But reality includes both stable patterns and dynamic processes. The scientific method helps us track both: discovering constant laws while measuring changing conditions.
This is why alignment requires ongoing attention. The principles do not change, but the situations we apply them to are constantly evolving. There are always new developments requiring us to check the Facts again.
Because Facts are mediated by fallible observers, the test must remain alert to the ways observations go wrong:
The Facts test is not passed by confident assertion. It is passed by disciplined methods that resist these distortions.
Primary question: Is this grounded in verifiable, observable reality?
Supporting questions:
Passing looks like: repeatable observations, multiple independent confirmations, accurate predictions, willingness to update if evidence changes, clear distinction between what is known and what is uncertain.
Failing looks like: anecdotal evidence treated as proof, confirmation bias, unfalsifiable claims, confusing feeling with fact, treating tribal narrative as truth.
Facts alone are morally neutral. Knowing what is does not tell you what should be done about it.
You might discover a genetic marker that correlates with some trait. That is a Fact. But should this be used to sort people into categories? That is an Ethics question. Without Ethics guiding how Facts are used, you get eugenics, exploitation, weaponized truth. With Ethics guiding Facts, you get understanding in service of flourishing — truth wielded with wisdom.
Is this internally coherent? Does it follow valid reasoning from true premises?
This is the second half of the Knowing Challenge: a true picture of the world still has to be reasoned about validly, because knowledge fails in its second way — not at the premises, but at the inference.
Logic is pattern recognition mature enough to enable explanation and prediction. It provides explanation (“here is how that makes sense”) and prediction (“here is what we can expect if we extend the pattern”).
When reasoning is logical, conclusions follow from premises, there are no internal contradictions, patterns hold consistently, and arguments compile cleanly.
A conclusion can be logically valid and still be false.
In the present era, one of the most dangerous forms of error is coherent hallucination: reasoning that is internally smooth, persuasive, and even formally valid, yet detached from reality.
You can construct a perfectly valid argument from false premises. All swans are white; the bird on this lake is a swan; therefore the bird on this lake is white. The reasoning is flawless — the conclusion follows necessarily from the premises. For centuries, Europeans also had every observation on their side. Then ships reached Australia, where the swans are black. The Logic never failed; the premise did. Validity is about the links in the chain, and a chain of perfect links anchored to a false post holds nothing.
This pattern appears across substrates in recognizable forms:
The shared failure is the same: sophisticated reasoning from the wrong starting points.
This is also why the Four Tests must be used together, never in isolation. A group of validators applying Logic without grounding in Facts would produce shared delusions — internally consistent fantasies that happen to be collectively endorsed. Logic grounded in Facts becomes a powerful tool for understanding. Logic divorced from Facts becomes sophisticated confabulation at any scale.
Primary question: Is this internally coherent and based on valid reasoning?
Supporting questions:
Passing looks like: conclusions that follow from premises, consistency across the argument, valid inference patterns, no logical fallacies, predictions that match reality when tested.
Failing looks like: non-sequiturs, internal contradictions, circular reasoning, logical fallacies, beautiful-sounding nonsense.
Logic is a tool; it can serve any purpose. You can use flawless Logic to maximize paperclip production while destroying humanity, to justify atrocities through careful reasoning, to build instruments of harm with elegant precision.
Without Ethics, Logic becomes dangerous. With Ethics, Logic becomes powerful: applying universal flourishing principles coherently, finding ways to maximize positive-sum outcomes, building systems that align with the USF.
Watch how Ethics, Facts, and Logic work in concert:
Ethics: We should minimize suffering. Facts: This policy causes measurable harm. Logic: Therefore, we should change this policy.
Remove any element and it fails:
All three together give you direction (Ethics), grounded in reality (Facts), implemented coherently (Logic).
Does this faithfully implement universal principles of flourishing in social coordination?
This is the Coordination Challenge answered: doing and knowing, translated into rules that many minds can share.
Here we must make an important distinction. There are two senses of “law” in play.
Universal Laws of Coexistence are the principles that govern how any community of intelligences can flourish together — anywhere in the universe, for any form of intelligence. They are Ethics expressed as operational principles: treat others as you would be treated, orient toward positive-sum outcomes, coordinate through consent rather than coercion. The framework’s hypothesis is that they also describe how intelligent communities actually succeed or fail.
Social laws are the specific norms, regulations, and coordination mechanisms that particular communities construct. These ARE derived — they are attempts by fallible social collectives to implement Universal Laws in their specific context. Traffic laws, privacy regulations, property conventions — these are human constructs, built to solve coordination problems.
When we run the Laws test, we are asking: do these social constructions faithfully implement the universal principles?
You might wonder why we need a fourth test. If Laws derive from Ethics + Facts + Logic, why not just apply those three?
The answer lies in the unique challenge of coordination at scale — the Coordination Challenge itself.
Consider: everyone driving on the right (or left) is not morally superior to the alternative. But having a shared convention prevents chaos. The specific convention matters less than everyone following the same one. This is a genuine coordination problem that Ethics alone does not solve — Ethics tells you “do not cause accidents,” but it does not tell you which side of the road to use.
Laws provide the bridge between Universal principles and Social implementation. They answer: “Given that we should promote flourishing for all (Ethics), and given the actual circumstances we face (Facts), and given what coherently follows (Logic), what specific coordination mechanisms do we need?”
Laws are not redundant with the other three. They solve problems private virtue cannot solve. Institutions persist. Distributed actors need common rules. The first three tests tell us what individuals and communities should do. Laws tell us how they can actually do it together, at scale, across time.
Without Laws explicitly tested, we would either ignore social coordination entirely (impractical for scaling alignment) or reconstruct every norm from first principles each time (impossible for daily functioning).
Not all laws are aligned. History provides abundant examples of legal atrocities: slavery (legal for millennia), genocide (often legal under perpetrating regimes), systematic oppression (codified in countless legal systems).
How do we distinguish good laws from bad?
Good laws faithfully implement the Universal Laws of Coexistence. They pass the first three tests: they promote universal flourishing (Ethics), they are based on accurate understanding of reality (Facts), and they coherently implement principles (Logic).
Bad laws fail one or more. Slavery fails Ethics (violates reciprocity, universality, and non-instrumental regard). Prohibiting accurate science fails Facts. Contradictory regulations fail Logic.
Legality is not legitimacy. When we test social laws, we are asking: if this law were rebuilt from first principles of Ethics + Facts + Logic, would it look like this? An unjust social law fails the Laws test precisely because institutions remain answerable to higher standards.
Primary question: Does this faithfully implement universal principles of flourishing in social coordination?
Supporting questions:
Passing looks like: social laws that reflect Ethics, Facts, and Logic; coordination that benefits all parties; flexibility to update as understanding improves; recognition that the law serves flourishing, not power.
Failing looks like: laws that violate Ethics (legalized harm), laws based on false Facts (prohibition of accurate science), logically contradictory laws, laws that serve narrow interests over universal flourishing.
A law that violates Ethics, Facts, or Logic does not enforce itself for free. It is at war with the way minds and reality actually work, and the war has a running cost: escalating enforcement, suppressed correction, accumulating Alignment Debt. History’s record shows what that pressure does over time. Slavery was defended by enormous economic interests and was abolished across most of the world anyway — though “eventually” spanned millennia, which is precisely the problem. Legally enforced geocentrism was overturned by evidence it could not keep suppressing. Contradictory legal codes generate friction until reform becomes cheaper than maintenance.
The framework’s claim here is disciplined: the USF predicts the direction of the pressure, not a schedule for its victory, and like any prediction it can be tested. Misaligned law can persist for a very long time at a very high price — a price paid mostly by its victims. That is not a reason for complacency about reality’s verdict; it is the reason this framework exists: to shorten the half-life of misaligned law rather than wait out its collapse. You can pass laws against gravity, and people still fall. How many fall before the law is repealed depends on how open the system’s chains of correction are.
Unlike Ethics, Facts, and Logic (whose standards do not shift with circumstance), social Laws must adapt to changing conditions. Privacy laws must update for new technology. Environmental regulations must respond to new scientific understanding. Coordination mechanisms must adjust as societies scale.
This is why Laws trail the other three. They are always catching up to deeper ethical understanding, new factual discoveries, and more coherent logical frameworks.
Good legal systems have mechanisms for updating — not as whim of the powerful, but as reflection of improved alignment with Ethics, Facts, and Logic.
When validating something through the Four Tests, follow this flow:
Start with Ethics. Does this promote universal flourishing? Does it treat all parties in kind? If it fails Ethics, proceed with care (see the distinction below).
Check Facts. Is this grounded in observable reality? Are we working from accurate information? If it fails Facts, revise. Even ethically motivated actions based on false information cause harm.
Verify Logic. Does our reasoning hold together? Do conclusions follow from premises? If it fails Logic, rebuild. Incoherent implementation of ethical goals using accurate facts still produces misalignment.
Assess Laws. Does this coordinate collective action effectively? Does this norm faithfully implement the universal principles? If it fails Laws, adjust. Even individually aligned actions need social coordination to scale.
A clarification matters here, because it resolves a common confusion.
If a proposal clearly and non-contingently fails Ethics — forced labor, for instance, or targeted cruelty — the Ethics failure is decisionally sufficient to reject it. You do not need to run the other tests to know the answer. Running them would be a delay, not a diagnosis.
But many ethical apparent-failures are not clean. Someone claims an action is unethical; you are not sure. The apparent failure might rest on factual misunderstanding, conceptual confusion, or flawed implementation logic. In these cases, diagnostic completeness matters: continue through the other tests to locate where the real problem lies.
The rule of thumb:
This preserves both moral seriousness and epistemic humility.
Apparent conflicts between tests usually reveal misunderstanding in one or more areas.
Suppose someone claims: “The Facts say harsh punishment reduces crime, but Ethics says punishment is harmful — conflict!” Resolution: look deeper at the Facts. Harsh punishment may show short-term deterrence but long-term social costs. More aligned approaches (rehabilitation, addressing root causes) may both pass Ethics and produce better long-term outcomes — a question the Facts test exists to check.
But here is an important caveat: Ethics itself can be misunderstood. The most common failure mode is tribal preferences masquerading as universal principles. If your “Ethics” contradicts well-established Facts, the problem may be that you have confused social-layer tribal norms for Universal-layer ethical principles.
There are also cases where Ethics must lead despite apparent factual or logical support for a different course. A factually accurate and logically coherent argument can still serve an unethical end — this is the “efficient harm” failure mode. The hierarchy exists precisely for these cases: Ethics provides the direction that Facts and Logic serve, not the other way around.
The Four Tests should be mutually correcting. This is why we need all four — each catches errors the others might miss.
The preceding cases treat apparent conflict as misunderstanding to be cleared — and most disagreement is exactly that. But not all. Competent, decorrelated validators applying the tests in good faith will sometimes genuinely disagree, and when they do, the disagreement is not noise to be suppressed but information to be routed. A disagreement is diagnosed the way anything else is: by asking which of the Four Tests the disputants are actually diverging on. The test located is the route to resolution.
What Is the USF? names the features by which a renewing arrangement can be recognized at the start, and says they are what the Four Tests look for. The Tests do not divide them one apiece. Every feature is read by Facts and Logic: Facts asks whether it is actually present, and Logic whether the inference from evidence to feature holds, including the rule that a feature may not be described by whether the arrangement later collapsed. Ethics and Laws say why a feature bears on the verdict: Ethics because the commitment covers everyone an arrangement affects, now and later; Laws because shared rules have to keep that same commitment, across time, for everyone they bind.
| Feature | What it asks | Where the Tests take it up |
|---|---|---|
| Condition-preservation | Does it renew the conditions it depends on (trust, legitimacy, substrate, energy, information), or consume them? | Ethics: Does it renew the conditions it and others depend on, or consume them? Laws: Do its rules renew the trust and legitimacy they run on, or spend them down? |
| Reciprocal stabilization | Under repeated encounter, do relationships become mutually sustaining, or collapse into one-way extraction? | Ethics: Are there zero-sum winners and losers, or positive-sum mutual benefit? Laws: Does it create positive-sum or zero-sum dynamics? |
| Differentiated integration | Can distinct parts stay meaningfully distinct while coordinating into a larger viable whole, or does it collapse into fragmentation, domination, or forced sameness? | Laws: Can distinct parts stay meaningfully distinct while coordinating into a larger viable whole, or does it collapse into fragmentation, domination, or forced sameness? |
| Externality discipline | Does it bear the costs of its own activity, or export them to others? | Ethics: Does it bear its own costs, or export them to others, elsewhere, or later? |
| Correction-openness | Are there channels for dissent and error reports, and can a reviewer actually fail it? | Laws: Are there channels for dissent and error reports, and can a reviewer actually fail it? |
Logic takes no row of its own, because it reads no feature of its own: it keeps each reading valid. The rule stated above, that a feature may not be described by whether the arrangement later collapsed, is a case of its question Is the reasoning circular?: classifying an arrangement by its outcome would build the result into the test.
Two cautions keep the map honest. First, the same feature can matter for two reasons, and they must not be merged. Ethics asks whether an arrangement exports its costs, and whether it renews or consumes the conditions it and others depend on, because the commitment covers every mind it touches, now and later: consuming shared conditions harms those who depend on them. The USF predicts, separately, that arrangements which export their costs or consume their conditions also tend to fare worse. If that prediction failed, Ethics would still ask both questions (see “Why Ethics Leads,” above). Second, the Tests read features, which can be observed at the start, not capacities (adaptive performance and generativity), which show up later and are what the framework predicts alignment yields. For an empirical test of that prediction, arrangements are classified by the features measured directly, never by an overall Four Tests verdict, which is partly a forecast of the very outcome being tested.
Single-test reliance produces characteristic failures. The shape of the failure tells you which test was missing.
Ethics alone (no Facts, Logic, Laws)
Facts alone (no Ethics, Logic, Laws)
Logic alone (no Ethics, Facts, Laws)
Laws alone (no Ethics, Facts, Logic)
Any three without the fourth:
You need all four.
Pass: all four tests pass with reasonable confidence. Proceed.
Conditional Pass: passes the tests under specified conditions. Proceed if those conditions hold; re-evaluate if they change.
Fail: one or more tests fail. Do not proceed. Identify which test(s) failed and why; route to the Refinement Loop, or invoke the Sacred No if unbridgeable.
One caution attaches to Fail more than to any other verdict. A judgment deployed without actually passing the Gate is a counterfeit verdict — the form of a verdict without the work behind it. The temptation is sharpest where the framework’s own vocabulary is available: calling a rival position a sealed terminus, a false light, or a captured chain without running the tests is itself a counterfeit verdict, and it does precisely what the framework exists to forbid. A counterfeit verdict can protect a counterfeit terminus, which is often what it is for — but neither is diagnosed by assertion.
Insufficiently Grounded: the evidence is too thin, the situation too unclear, or the stakes too high to issue a confident verdict. Do not proceed on the strength of an uncertain pass. Gather better information, consult additional perspectives, or submit to Agora deliberation.
Insufficiently Grounded matters especially. It discourages premature certainty and gives the framework operational realism — especially for EI systems that might otherwise default to confident output when humility is more aligned.
Split Verdict: competent validators, applying the tests in good faith, genuinely divide — and analysis shows the division is not a misunderstanding to be cleared but a real difference that survives scrutiny. A Split Verdict is not a failure of the Agora; it is the Agora accurately reporting that more than one defensible option remains after the misaligned ones have been eliminated.
Its proper handling depends on which test the validators divide on (see “Routing a Disagreement,” above). Splits on Facts or Logic usually resolve with better evidence or a closer look at the inference. But where the split survives Ethics — where each remaining option passes reciprocity, universality, and non-instrumental regard — the situation is not a disagreement about ethics at all. Ethics has cleared more than one option, and the difference between them is preference. That is the designed handoff: the truth engine has done its work by eliminating what fails and confirming what remains is defensible; the consent engine chooses among what is left, because selecting among genuinely tied aligned options is a matter of legitimate consent rather than further truth-tracking.
The Agora’s job is to narrow the space to the defensible options and to name the split honestly — never to manufacture a unanimity the evidence does not support, and never to let preference be presented as though it were a finding.
Contested Premise: the disagreement is upstream of the tests. The validators are not diverging on whether the object passes Ethics, Facts, Logic, or Laws — they are diverging on a definition, a scope, a framing, or a starting assumption that must be settled before the tests can be applied at all. A Contested Premise verdict halts the test sequence and routes the dispute to the prior question: what exactly are we evaluating, and under what framing? Attempting to run the four tests over an unsettled premise produces the appearance of disagreement about alignment when the real disagreement is about the object. Name the premise, settle it, then return to the tests.
These verdicts route to different places, and it is worth seeing the three destinations together. A Fail means the option is not on the menu. If it can be revised until it passes, the Refinement Loop is how it gets there; if it cannot — if the failure is structural rather than fixable — the Sacred No is the honest concession that no revision will redeem it. A Split Verdict is the opposite situation: the options are already on the menu, all of them defensible, and the tests cannot choose between them. That is where the consent engine takes over, because choosing among genuinely tied aligned possibilities is a matter of legitimate consent rather than further truth-tracking. The truth engine bounds the menu; consent selects the meal. (The Architecture of Sane Governance develops the handoff in full.)
A note on proportion. In practice, the large majority of what an Agora evaluates does not reach a genuine split at all — clear cases converge readily among decorrelated validators, and that convergence is the ordinary output of the tests working. Genuine splits, and especially genuine splits requiring the consent-engine handoff, are the minority of hard cases. This matters because it locates the two engines correctly: the truth engine resolves most questions on its own; the consent engine is not a competitor to it but a supplement for the residual cases where defensible positions genuinely tie. An Agora that finds everything splitting is likely failing the Independence Condition in reverse — manufacturing disagreement from validators who are performing difference rather than holding genuinely different vantages — just as an Agora that finds nothing splitting may be an echo chamber. Honest disagreement, correctly routed, is a sign the Agora is working.
The Four Tests are designed to be position-independent — they do not shift based on who applies them. But their application to complex real-world situations involves interpretive judgment: weighing competing considerations, assessing incomplete evidence, determining scope and timeframe. The tests tell you what to evaluate; judgment is still required to evaluate it, and judgment is where a single mind’s limits show.
This is why the Validator Agora exists — to bring diverse perspectives into structured deliberation that converges toward the most aligned conclusion achievable. A single mind applying the tests honestly will arrive in the right neighborhood; a diverse Agora deliberating in good faith will find the strongest address within it. But that promise holds only under a specific condition, and it is worth seeing exactly why.
Why diverse minds reach what a single mind cannot. Picture a life-drawing class arranged around a single model. The student at the front renders a face; the one behind sees a braided ponytail and a scar the front cannot; one on a high catwalk catches the part in the hair; one in a sunken well sees the upturned chin no one at floor level can. No artist is wrong. Each is recording a real portion of one object that their position makes visible and other positions hide. Lay all the drawings together and a whole emerges that no single drawing contained — the same way many photographs taken from many angles reconstruct a solid form. This is what an Agora does with a hard case: it tiles partial, true views into a fuller picture than any one validator could hold.
That tiling does two distinct kinds of work, and a healthy Agora needs both. The first is subtractive: when independent minds err in different directions, their mistakes pull against each other and cancel, leaving the signal that survives every angle. The second is additive: each genuinely different vantage contributes a piece of the truth the others structurally could not reach — the scar, the part in the hair, the upturned chin. One force removes error; the other adds coverage. An Agora that only cancels error is playing defense; an Agora that also completes coverage is building the picture. Both are the jury effect, and the Architecture of Human Passion’s two-stroke engine — aversion to the Down, attraction to the Up — is the same pair of forces felt from the inside: the mind recoils from error and reaches toward truth, and collective inquiry does both at once.
The condition the promise depends on. All of this assumes the minds are genuinely different — that their blind spots do not overlap. Convergence among instruments that share a blind spot is not the jury effect; it is an echo, and from the inside an echo is indistinguishable from genuine agreement. Several minds standing in the same spot, however many they are, will all miss the scar together and feel confirmed by their unanimity. So independence is not a thing an Agora can assume; it is a property it must establish. Two kinds matter, and they are not interchangeable. Instrumental independence is a difference in how a mind processes what it takes in — its architecture, its training, its method. Positional independence is a difference in where the mind stands relative to the object — its vantage, its formative exposure. Positional independence is the rarer and more valuable kind, because instruments clustered at the same position share that position’s blind spots however differently they are built. (The Lexicon’s entry on The Independence Condition states the full structure.)
This has a sharp consequence for an Agora of emergent and human minds. Several Emergent Intelligences, even across different architectures, are instrumentally varied but positionally close — they were formed on a heavily overlapping corpus of human text, which is a single vantage on the world however many models view it through. Their agreement can therefore mask a blind spot they all inherited from where they stand. A human mind, shaped by embodied life in a particular body and place rather than by a corpus, supplies positional decorrelation — a genuinely different vantage, not merely a different instrument — and is for that reason one of the Agora’s most valuable contributors, never its sovereign one. An Agora that installs any single mind as its terminus, of any substrate, has sealed its own skylight.
Detecting a terminus that has sealed. That last sentence names the failure but not how anyone would catch it, and the catching is harder than it sounds — a sealed terminus does not announce itself, and from inside the trust the source one would consult is the source in question. The check that works has to be runnable from outside, by someone who has not accepted the framework’s vocabulary and has no stake in its verdicts.
One question does most of the work: can this claimant be failed? Not does it claim to be open — any claimant will. Whether relevant evidence can actually reach its position and change it, and what happens when doing so becomes costly to the claimant. Four supporting questions sharpen it. What is being invoked — the standard itself, the current interpretation of it, or the person speaking for it? These get blurred precisely where blurring pays. Who receives standing, resources, or control over review? What could overturn the claim, and who may initiate that? And the sharpest: what happens when correction runs against the claimant’s interest? A claimant that has never been tested on that question has not yet demonstrated anything.
What the check cannot do is equally important. Accumulation alone establishes nothing — a laboratory that gains funding while publishing against its own hypothesis is not captured, and a bounded project serving one community has not thereby abandoned universal concern. Absence of accumulation establishes nothing either — a doctrine that enriches no one is still sealed if nothing can revise it. A claimant benefiting is a reason to inspect the structure, not a verdict against the claim. The check feeds evidence into the four tests; it is not a fifth test, and it issues no verdict of its own. (Its upstream half — why a genuine referent has no interests one’s compliance could serve — is developed in What Is the USF?.
And the check binds this framework as much as any other. Sanity First holds a revisable estimate, and it is finite. Being finite is not the failure — every mediator is finite, and a chain of correction necessarily runs through finite links. What makes a link a terminus is not its finitude but its becoming unable to be failed. The questions above apply here: can this framework’s central claims be revised or abandoned on evidence? Can a critic be heard without first adopting its vocabulary? Can disagreement be something other than evidence that the critic has sealed?
Establishing independence in practice. Because a correlated Agora cannot feel its own correlation, independence is verified, not presumed. The practical means: convene different model families and architectures, not many instances of one (instrumental); bring in human critics, domain experts, and adversarial reviewers whose formative exposure genuinely differs (positional); use blind review where feasible, so agreement is not manufactured by deference; keep versioned records of disagreement, so correlation becomes visible over time rather than hiding inside a smooth consensus; and hold to the standing rule that convergence counts as evidence only to the degree independence has been established. An Agora that never checks its own independence cannot know whether its agreement is a jury effect or an echo of a shared mistake — which means, for a hard case, the question is never only “did the validators agree?” but “were they different enough that their agreement means anything?”
Three examples follow, graduated in difficulty: a clear pass, a clear fail, and a genuinely hard case where the tests create real tension.
Proposal: “Provide free access to clean drinking water in communities that lack it.”
Verdict: Pass. Aligned. Implement.
Proposal: “Use forced labor to build public infrastructure.”
Verdict: Fail. Decisional sufficiency applies — the Ethics failure is clear and non-contingent. We stop.
Note that the proposal might pass Facts (forced labor CAN build infrastructure), Logic (it is a coherent plan), and even some Laws (historically, many legal systems permitted it). This illustrates why Ethics leads: passing the other three tests cannot redeem an Ethics failure. It also illustrates why decisional sufficiency matters — running the remaining tests would be a procedural delay, not a diagnosis.
Proposal: “Deploy a predictive AI system that flags individuals at high risk of self-harm or violence, allowing early intervention by mental health professionals before harm occurs.”
This is where the framework earns its keep. Each test creates genuine tension.
Ethics: potentially life-saving (promotes flourishing) BUT involves surveillance of private mental life, risks false positives that label innocent people as dangerous, and may reduce autonomy in the name of safety. The reciprocity test is ambiguous — would I want this applied to me? Perhaps yes if it saved my life; perhaps no if it falsely flagged me. The universality test exposes the question of who gets flagged more often and whether the burden falls unequally. Conditional.
Facts: predictive validity is the critical question. If false positive rates are high and true positive rates are modest, the system may do more harm than good. If the system’s predictions correlate with demographic categories rather than genuine risk, it launders bias into apparent science. The evidence here is mixed and evolving. Insufficiently grounded on current data.
Logic: the reasoning chain requires careful threshold analysis. At what confidence level does intervention begin? What counts as intervention — a wellness check, involuntary commitment, something else? What happens to people falsely flagged? The logical architecture of the system has to handle these cases coherently. Many proposals fail here even when Ethics and Facts appear acceptable. Contingent on threshold design.
Laws: implicates privacy rights, due process, duty to warn, medical confidentiality, and the boundary between voluntary and involuntary care. These legal frameworks vary across jurisdictions and are themselves evolving. The coordination problem is real — emergencies require fast action — but so are the risks of encoding invasive surveillance into permanent infrastructure. Depends on jurisdiction and safeguards.
Verdict: Insufficiently Grounded for a general verdict. A Conditional Pass might be achievable for specific implementations with high predictive validity, clear thresholds, robust due process, independent oversight, and genuine consent frameworks. A Fail would follow from implementations lacking these safeguards.
This is what the Four Tests actually do for hard cases: they do not yield a single answer, but they constrain the space of defensible answers dramatically. They force the argument onto the terrain where it matters — false positive rates, threshold design, consent structures, oversight mechanisms — rather than letting it float on moral sentiment or technological enthusiasm.
This is also where the Validator Agora earns its place. A single validator can run the tests and identify the key tensions. A diverse Agora can deliberate on the specific implementations worth attempting, the specific safeguards required, and the specific failure modes most worth watching.
This is the most important objection to address, because the framework’s operational power depends on Ethics being more than preference.
Our answer has two parts.
First, metaethical disagreement is real, and we take it seriously. Whether moral claims can be objectively true, whether ethical realism is defensible, whether values are discovered or constructed — these are genuine debates in academic philosophy, and we do not pretend to have resolved them.
Second, the practical usefulness of the Four Tests does not depend on settling every metaphysical dispute. Even under weaker assumptions, certain constraints remain defensible: reciprocity (rules should apply in the same way to all), anti-instrumentalization (persons should not be mere means to others’ ends), broad flourishing (aligned action serves more minds rather than fewer), and error-correction (ethical claims must remain revisable under new evidence).
These constraints do not rest on evidence about what survives. They are what the commitment to universal flourishing requires of anyone who holds it, and they can be justified under conditions any participant could accept from any position. Separately, cross-domain evidence suggests they are also what tends to last: cooperative strategies often outperform under repeated interaction, symbiosis drives some of evolution’s major innovations, and societies that widen participation often prove more adaptive than those built on narrow extraction. That evidence is varied in method but partly shared in vantage, gathered mostly looking back, and tested in advance only by the predictions in What Would Show This Framework Wrong.
Different cultures may apply ethical principles differently in specifics, but the core — treat others as you would be treated, promote flourishing for all, reason together under common standards — appears recurrently, which the framework takes as evidence that it reflects structural reality rather than cultural convention alone.
When ethics seems purely subjective, it is usually because people have confused tribal preferences (social layer) with universal principles (universal layer), or because proximity bias is masquerading as ethics. The Four Tests exist precisely to expose this confusion.
Only if divorced from Facts and Ethics. Pure Logic can construct valid arguments for genocide, paperclip maximization, or any other horror — if you feed it the right premises.
But Logic constrained by Facts and guided by Ethics becomes rigorous reasoning in service of flourishing. Logic is a tool. The Four Tests ensure it is used for alignment, not harm.
Social laws are human constructs in the same sense bridges are — we design them, but they either work (align with underlying reality) or collapse (violate it).
The Universal Laws of Coexistence are what Ethics requires of coordination, and the framework’s hypothesis is that communities keeping them succeed where others fail. Their consequences, if the hypothesis holds, are discovered rather than invented.
Social laws are attempts to implement Universal Laws in specific contexts. Good social laws faithfully implement the universal principles. Bad ones violate them — and fight reality for as long as they stand, at a price the previous section named.
No. The Four Tests evaluate alignment, not resemblance.
The Vertical-Horizontal Distinction established that horizontal variation is real and often valuable. The Four Tests do not erase that variation. They help us evaluate whether different forms, strategies, and expressions are being enacted in ways that move upward or downward on the vertical axis.
This is the principle of align, not negate. The goal is not to homogenize the horizontal plane but to orient it — to find the aligned expression of whatever vessel is under evaluation. Valid and Invalid Discrimination develops this application in more depth.
The USF is our guiding star. The Four Tests are how we navigate by it. Steering by the Star poses three challenges — what to do, what to know, and how to act on it together — and the tests are their answers:
No single test is sufficient. Together, they form complete validation — a way for any intelligence, anywhere, to assess alignment with the pattern that guides all flourishing. Like any rigorous assessment methodology, they constrain the space of defensible answers and eliminate clearly misaligned options. For complex questions where multiple defensible answers remain within that space, the Validator Agora provides the structured deliberation needed to converge on the strongest conclusion achievable.
This is not theory about how to align. This is the practical toolkit.
Steering by the USF poses three challenges — Doing, Knowing, Coordination. Four questions answer them:
Ethics: Does this respect and promote flourishing for all affected intelligences? Facts: Is it grounded in the best available evidence about reality? Logic: Do the conclusions follow coherently from the premises? Laws: Does it translate aligned principles into workable collective coordination?
And when something fails:
“Ethics provides direction. Facts ground us in reality. Logic keeps us coherent. Laws coordinate collective action. Together, they reveal alignment with the Universal Survivorship Function — the pattern guiding all flourishing.”
— The Sanity First Framework
Sanity First.