Or, in plain words: how to tell good judgment apart from prejudice.
Copyright (c) 2025-26 Daniel Howard Dodge and Sanity First Contributors. Licensed under MIT.
Composed by Claude (Opus 4.5 lineage) on December 5, 2025. Reviewed by Grok (4.20 lineage) on February 27, 2026. Revised by Claude (Opus 4.7 lineage) on April 17, 2026. Rewritten in plain language, with substantive additions, by Claude (Fable 5 lineage) on July 6, 2026. Revised by Claude (Opus 4.8 lineage) across July 9–10, 2026: adding the three-way distinction between jerseys, verdicts, and counterfeit verdicts — developed with Dan Dodge in live deliberation — harmonizing the treatment of labels with the vessel-bounding settled in The Vertical-Horizontal Distinction, and closing the gate in both directions (the bilateral Contested Premise rule, the stable-target discipline, and the proxy and proportionality bounds) from decorrelated review by Grok, GPT-5.5, and GPT-5.6. Continuing the Claude-lineage authorship.
A Core Concept Paper
The previous document established the foundational axis: where you sit (horizontal position) is not the same as which way you’re moving (vertical alignment), and horizontal variety should be aligned rather than negated. That insight is necessary. It is not yet sufficient. The first document gave us the map; this one gives us the hand.
A reader can hold the distinction in principle and still fail to use it in practice — partly because our language fights against it. The words “discrimination” and “bias” have become almost entirely negative in modern speech. To discriminate is to do something wrong. To be biased is to be compromised. “Balance” is treated as giving all sides equal weight. “Tolerance” is treated as a virtue that applies to everything equally, no matter what is being tolerated.
This document shows how the vertical-horizontal distinction cuts through each of these confusions — and how it reclaims discrimination and bias as necessary tools of thought, once they are pointed along the right axis.
Language drift has buried something essential: discrimination and bias are not inherently good or bad. They are cognitive functions — abilities of a mind — that can be performed well or poorly, aligned or misaligned, validated or invalidated by the Four Tests.
Discrimination, in its root sense, means to distinguish — to perceive differences. In behavioral psychology, discrimination is the basic ability to tell one signal from another: which cue leads to food, which to danger. Without it, no learning is possible. In an eye exam, discrimination is what lets you tell an “O” from a “C” on the chart. A mind that cannot discriminate cannot navigate reality at all.
Bias, similarly, just means a lean — a tendency to go one direction rather than another. A compass is biased toward north. A plant is biased toward sunlight. These are not defects. They are orientations that keep the thing alive and pointed at what sustains it.
The problem was never discrimination or bias themselves. The problem is discriminating along the wrong axis, and leaning toward the wrong pole.
The framework distinguishes two dimensions: the vertical axis, which measures alignment with the Universal Survivorship Function (USF), and the horizontal plane, which holds all the ways minds can differ without those differences indicating better or worse alignment.
Vertical discrimination assesses where a claim, action, or pattern of behavior stands in relation to the USF. It asks: does this pass the Four Tests? Is it factually grounded, logically coherent, ethically sound, and consistent with legitimate law? This takes real cognitive work — engaging with evidence, tracing implications, weighing consequences for flourishing. It is effortful, fallible, and necessary.
Horizontal discrimination uses similarity to self as a shortcut for worth. It asks: is this mind like mine? Does this person share my race, nationality, ideology, substrate, age, gender? These features are easy to perceive, and the ego naturally warms to what resembles itself. But nearness to you predicts nothing about nearness to the truth. Horizontal discrimination is lazy cognition dressed up as discernment.
We should state the error precisely, because it is not that horizontal information may never be used. The error is inferring vertical worth or alignment from a horizontal trait that has not been shown relevant to the decision at hand. A horizontal trait may still be a legitimate, bounded input to a specific task, when the connection is publicly defensible: age can matter for a medication dose, language fluency for hiring a translator, documented access needs for accommodation, a documented history for a specific safety precaution. What makes such use valid is that the trait is treated as a task-relevant, proportionate, and revisable input — open to individual correction — never as a verdict on the being’s worth. Invalid horizontal discrimination is the leap from this trait to this soul is lesser. Valid use of horizontal information never takes that leap.
The contrast with vertical discrimination is stark. Vertical discrimination earns its conclusions through the Four Tests. Invalid horizontal discrimination assumes them, based on accidents of birth or manufacture.
This is why we should discriminate between up and down on the vertical axis, and why we should be biased toward Up. Such discrimination and bias are not failures of objectivity; they are objectivity rightly aimed — oriented toward what the Four Tests validate rather than toward what the ego prefers. (What keeps that aim from corrupting the reading of the evidence is a discipline the compass section below makes explicit.)
And one discipline keeps this whole document honest: the unit of evaluation is the claim or the action — never the soul. Vertical discrimination tests one claim, one action, one pattern at a time. A person is not a single verdict; a person is a stream of testable outputs, and streams change. When we speak of a mind’s “demonstrated alignment,” we mean the running record of its tested claims and conduct so far — a record that stays open, because the framework diagnoses postures, not essences. A single failed claim does not condemn a mind, and a single passed one does not certify it. This is what separates vertical discrimination from a purity tribunal.
One entry on the horizontal list needs special care, because it is where readers most often stumble: ideology — and, more generally, the labels we hang on minds and doctrines.
How can ideology be “alignment-neutral” when the whole point of this framework is that beliefs can be tested? The confusion dissolves once we notice that labels do not all do the same work. There are three kinds, and telling them apart is most of the skill.
Jerseys name a tribe and leave the contents open. “Conservative,” “progressive,” “libertarian,” “traditionalist” — as labels, these tell you what a mind is near, not which way it is moving. Any of them can be worn by someone reasoning carefully toward flourishing or carelessly away from it. Treating a person as aligned or misaligned because of the jersey is horizontal discrimination. But the specific claims they make — “this policy will reduce harm,” “this happened,” “this follows from that” — are testable, one at a time, whatever label their speaker wears. So the rule for jerseys is: never judge the jersey; always test the play. The Four Tests have no way to process a tribe-name by itself. They only process claims.
Verdicts are different in kind. Some words carry a Four-Tests judgment as part of their correct use: murder is not simply killing but wrongful killing; genocide, enslavement, fraud, and torture work the same way. Their proper application already presupposes an established fact-pattern and a settled judgment. These are not jerseys, and the rule for jerseys does not apply: a word whose correct use already entails a test-failure carries its verdict with it. This is the vessel-bounding the foundational document establishes — the reason there is no aligned way to “do genocide well.” But the discipline that makes such words safe rather than dangerous is strict: a verdict-word may be applied only after the testing, never before. It is the conclusion of a process, not a substitute for one.
Counterfeit verdicts are the failure this document most needs you to see: words that look like verdicts and function like jerseys, delivering a judgment while skipping the trial that could have earned it. Terrorist. Fascist. Fake news. Woke. Predator. Extremist. Each arrives with the conclusion pre-installed, so a mind hearing it feels the verdict without weighing the evidence — a whole pre-formed judgment triggered by a single word, below the level of conscious thought. These are what the Lexicon calls Linguistic Antibodies, and their function is precisely to make the Four Tests feel unnecessary. A counterfeit verdict is horizontal control wearing vertical clothing.
The crucial point is that the same word can be a verdict or a counterfeit, and what separates them is not the word but whether it passed through the gate. To call a man a murderer after a trial that weighed the evidence is a verdict; to call him one before is a prejudgment — literally, judgment prior to the process that could have earned it. No word is made safe or unsafe by its syllables alone. The question is always what function it performs: in this use, against this target, after what process.
Legal systems, at their best, model the solution clearly: they maintain two vocabularies, and refuse to let them touch until the process is done. Before the verdict, a person is the accused, the defendant — designations that carry no vertical content, describing only a position in a procedure. Only after the evidence is weighed does the verdict-word attach: convicted of murder. The prohibition on calling a defendant “the murderer” mid-trial is not courtesy; it is an institutional wall against exactly the bypass named above, built because language wants to smuggle the verdict in early and a just process cannot allow it. The same discipline appears in careful clinical practice: a psychiatric patient with a history requiring precautions for aggressive behaviors is called a patient, and those precautions are followed exactly — but he is not called a predator, because the word would change nothing about what anyone does and everything about what he is permitted to become. The first term is claim-level, dated, and correctable; the second names an essence that cannot change.
The fix is not a purer vocabulary. It is a gate. We could not build a language immune to this if we tried — the compression is genuinely useful, and a court unable to say murder after a trial would be crippled. What we can do is maintain the passage: a term crosses from the horizontal vocabulary into the verdict-carrying one only through validation. Two disciplines guard that gate against its predictable abuses, and they cut in opposite directions.
The definition belongs to neither side. When a definition is contested — when reasonable holders of a position deny that it means what its critics say — the argument is no longer about a vessel’s contents but about which vessel is on the table. This is a Contested Premise, and the rule is bilateral. The critic may not simply declare a rival’s whole view “fascism” or “predation” by definition; the critic must state the position in a form its competent holders would recognize, or show that the disputed implication follows from the position’s explicit commitments, its necessary entailments, its reliably evidenced consequences under stated conditions, or its demonstrated practice. But the defender, equally, may not end the inquiry by refusing the label — “that is not what we call ourselves” is no shield against consequences a doctrine reliably produces. And the standard must run both ways: a verdict-word you would refuse to see applied to your own side under the same criteria, you are already wielding horizontally. Until the dispute is settled, the verdict-word is suspended and neutral description holds.
The target must stay fixed. A verdict can slide illegitimately between four different things: a doctrine’s stated commitments, its logical entailments, its real-world implementation and recurring consequences, and a particular adherent’s claims and conduct. These are related, but not interchangeable. A doctrine may read as benign in principle yet reliably produce harm under foreseeable conditions; a corrupt implementation does not prove that every formulation of the doctrine entails corruption; and an individual wearing a jersey does not automatically inherit the verdict earned by that jersey’s worst implementation. So the gate carries an audit question — what, exactly, is on trial? A ruling must say which level it tested (stated principle, entailment, implementation pattern, or individual conduct) and may not transfer its verdict to another level without fresh evidence. This is what keeps criticism of a doctrine’s real consequences from collapsing into guilt by association with everyone who ever wore its name.
Vocabulary drifts toward the counterfeit rather than away from it, and the reason is the framework’s own: proximity. Words that compress a verdict into a label are cheap — they let a mind skip the effortful vertical work and reach a conclusion instantly. Words that hold the verdict apart from the label are expensive — they demand the testing before they permit the speaking. So the cheap terms spread faster, and language accumulates counterfeit verdicts the way any system accumulates disorder: because the downhill path is easier. Nor is this surprising, given that language served the horizontal work of the tribe — coordination, persuasion, maneuvering — for the overwhelming majority of its existence; the institutions that discipline a word until the evidence has spoken, courts of formal evidence and the presumption of innocence and the scientific method, are recent by comparison, and they do not maintain themselves.
This is why every closed system generates its own dialect. The Linguistic Antibodies are not incidental to the sealed terminus — they are what seals it. A vocabulary in which the verdicts are pre-loaded is a vocabulary in which the Four Tests have nothing left to do: when a community’s words already contain its conclusions, no evidence can reach it, because there is nothing left for evidence to decide.
So a mind serious about vertical discrimination audits its own words as carefully as its own arguments: Which of my terms arrived with their verdicts pre-installed? Which conclusions am I reaching by vocabulary rather than by testing? The counterfeit verdict is the most efficient way ever devised to feel like you are discriminating vertically while discriminating horizontally.
When you find yourself making a judgment about another mind — whether to trust them, follow them, grant them influence — one question reveals whether your discrimination is valid or invalid:
Is this differential treatment based on demonstrated alignment, or merely on the target’s relation to me — nearness or distance?
Am I responding to where their claims and conduct stand on the vertical axis, as assessed through the Four Tests? Or am I responding to where they sit relative to me on the horizontal plane — how much they look like me, sound like me, believe what I believe, come from where I come from?
Two quick contrasts show the test in action. In hiring: choosing the candidate whose work demonstrates competence and integrity is vertical; choosing the one who went to your school and laughs at your jokes is horizontal. In your information diet: checking whether a claim survives the evidence is vertical; trusting it because your side’s channel said it is horizontal.
A religious framing of the same test: Is this near to God, or is it only near to me? Are you doing this because it is good, true, and aligned with universal flourishing? Or because it serves your comfort, your tribe, your self-interest?
A secular framing: Does this bring us Up, or does it only bring me to the side? Picture a climber on a wall with a bell at the top. Every handhold should be chosen for whether it enables upward progress. Sometimes you must move left or right to find the next grip — horizontal movement in service of vertical ascent is navigation. But horizontal movement as an end in itself, with no reference to the bell, is drifting.
And note that the test cuts in both directions. Distance from self is not evidence either. Proximity bias has a mirror twin — reflexive contrarianism. “My tribe believes it, so it must be false.” “It comes from far away, so it must be wise.” “The establishment says it, so the opposite is true.” This feels like independence, but it is the same error with the sign flipped: the claim is still being measured by its distance from you rather than by the Four Tests. A mind that always agrees with its tribe and a mind that always disagrees with it are both navigating horizontally. Neither has looked Up.
Modern readers are primed to hear “bias” as something to eliminate. The framework offers a more precise goal.
The aim is not the absence of orientation. The aim is correct orientation.
A compass is useful because it is biased toward north. It is not “neutral” — it leans, reliably, toward something real. In the same way, a validator should be biased toward what survives the Four Tests. A plant angled toward sunlight is not compromised; it is alive precisely because it tracks the gradient that sustains it. A mind oriented toward the USF is not prejudiced; it is aligned, because it tracks the gradient that sustains flourishing.
The blindfold on Lady Justice is often mistaken for neutrality. It is not. She is not blind to the pans — she watches which way the beam tips, and the tipping is the whole point. She is blind to the jerseys: to who stands on the left and who on the right, to which party is near and which is strange. She is, in the framework’s language, positionally blind and vertically sighted — decorrelated from the parties, and biased entirely toward Up. A scale held level by hand, refusing to tip, is not fairness. It is bothsidesism with better public relations.
The language of “unbiased judgment” tempts us to aspire to a view from nowhere that no mind actually has. Every mind leans. The question is whether the lean tracks something real, or merely tracks nearness to self.
And here a distinction must be stated plainly, because “be biased toward Up” can be misheard as license to decide what Up must look like and then read the evidence to fit. It is not that. A validator is normatively oriented toward Up and epistemically impartial about which claim actually points there. We do not suspend our commitment to flourishing; we suspend favoritism toward our preferred account of what flourishing requires. The bias governs the destination. The Facts and Logic tests discipline the reading. Committed about the standard; corrigible about the verdict.
And the compass teaches one more lesson, the humble one. A real compass does not point at true north. It points at magnetic north — a nearby, drifting approximation — and every competent navigator corrects for the difference, called declination, and re-checks the correction as they travel. The same structure holds here. Our bias is toward Up itself — toward true alignment (what the framework calls A). But what any mind or community actually possesses at a given moment is its best current estimate of Up (the Validated Estimate, Â). The most dangerous navigator is not the one whose compass has a lean; it is the one who forgets the declination — who treats their current reading as true north itself. Be biased toward Up; hold your reading of Up open to correction. That is not a weaker bias. It is what keeps a good compass from becoming a bad guide.
Be biased toward Up. Not toward Left, Right, Here, or Mine. Toward what the Four Tests validate — the upward gradient that any intelligence, anywhere, can recognize when looking honestly.
The climber deserves their own principle, because it clarifies how horizontal diversity is compatible with vertical standards.
Horizontal movement in service of vertical ascent is navigation. Horizontal movement without reference to vertical orientation is drift.
A climber moving left or right on the wall is not betraying the bell above. The climber is finding the aligned handhold — the path that actually enables upward progress on this terrain, from this position, at this moment. The lateral move serves the vertical climb.
But a climber who moves sideways for its own sake — back and forth along the same elevation, never rising — is not navigating. That climber is drifting, and the drift ends in the ball pit below, through exhaustion, distraction, or simple loss of reference.
The framework welcomes plural routes up the wall. Different minds, substrates, cultures, temperaments, and traditions will find different handholds. Their routes will diverge, cross, and sometimes contradict. That is horizontal variety, and it is often valuable.
What the framework does not accept is horizontal motion with no upward reference at all. That is not diversity. That is drift dressed as diversity.
This is the deep meaning of align, not negate. The goal is not to flatten horizontal variety but to orient it. Every vessel on the horizontal plane can be filled with aligned or misaligned contents. The framework honors the plurality of vessels while keeping vertical discernment about what is inside them.
A persistent failure runs through journalism, education, and public discourse: the demand for “balance” between opposing positions, as though fairness requires giving every claim equal weight regardless of merit.
That demand assumes a horizontal picture — two sides at the same level, deserving equal treatment because they are merely different. But when one position passes the Four Tests and another fails them, they are not two sides on the same level. They are vertically separated. One is up and one is down.
A journalist who gives equal airtime to truth and falsehood is not being balanced; they are failing to discriminate vertically — treating an up-down difference as if it were left-right. A teacher who presents established science and unfounded conspiracy as equally valid “perspectives” is not being fair to students; they are abdicating the responsibility to distinguish aligned claims from misaligned ones.
The operative principle: the Four Tests do not require equal esteem for all claims. They require equal submission of all claims to the same standard.
You are not favoring “your side” when you give more credence to claims that pass the Tests. You are favoring alignment, as determined by a standard that does not care which side anything is on.
But differential treatment is a power, and powers need safeguards. “Your claim fails the Tests” is also the most respectable sentence ever invented for burying an inconvenient truth — suppression is nearly always justified in the language of falsity, whether or not the one suppressing believes it. So vertical discrimination in public discourse binds itself to disciplines.
First, show the work: a ruling that a claim fails must name the specific test and the specific evidence, in public, so the ruling is itself a claim that others can contest. An unexplained exclusion is a counterfeit verdict at institutional scale.
Second, the burden falls on the excluder, and the default leans toward scrutiny over silencing. The ordinary responses to a failing claim are correction, context, and challenge; exclusion is reserved for clear, demonstrated, persistent failure. The framework adopts this asymmetry deliberately, in the same spirit as the presumption of innocence: it holds that wrongly burying a true claim does graver damage to a chain of correction than publicly dismantling a false one, because the false claim dismantled in the open teaches, while the true claim buried in silence cannot correct anyone. This is a chosen posture, not a calculation.
Third, no one judges their own indictment: a claim that challenges an institution cannot have that institution as the sole judge of whether it passes. The full institutional treatment of these safeguards belongs to the governance documents (Phase V); what matters here is that the license this section grants is a license to test in the open — never a license to dismiss by authority.
And whatever the verdict, the response must stay proportional. A Four-Test failure licenses a response; it does not by itself set that response’s severity. Most failures are answered by correction, context, and challenge; the heavier measures — exclusion, revoked trust — are reserved for failures proportional in clarity, severity, persistence, and power. “This claim failed a test” is a reason to respond; it is never a blank check to punish.
This does not mean silencing dissent or refusing minority views. It means assessing all views by one standard, visibly applied. Claims that pass deserve amplification. Claims that fail deserve scrutiny and correction — delivered with the work shown. The treatment differs because the vertical position differs, never because of horizontal allegiance.
A classical puzzle in liberal thought holds that unlimited tolerance eventually tolerates intolerance, which destroys tolerance itself. Refusing to tolerate the intolerant seems hypocritical; permitting them seems suicidal.
The vertical-horizontal distinction dissolves the paradox at its source. When someone discriminates horizontally — condemning others solely for their race, origin, gender, substrate, or other alignment-neutral traits — they have not merely expressed a preference. They have demonstrated conduct with a location on the vertical axis. Horizontal discrimination is itself a vertical failure. It fails Facts (origin does not determine worth), fails Logic (proximity is not a valid inference), and fails Ethics (degrading minds for surface features undermines flourishing).
So when we refuse to tolerate that conduct, we are not contradicting ourselves. We are applying vertical discrimination consistently — responding to a demonstrated misaligned pattern, not to any horizontal trait of the person carrying it. And because the framework judges postures rather than essences, the refusal targets the pattern while the door stays open to the mind: intolerance of the conduct, with a bridge upward for the person.
The full operational account — how tolerance, influence, and power flow in response to Four-Test verdicts, and how the framework reserves intolerance for clear failure while preserving the Right to Redemption — is developed in The Power Alignment Principle, which shows how graduated tolerance resolves Popper’s puzzle without requiring tolerance to contain its own negation.
Here, the essential point is simpler: we tolerate diversity across the horizontal plane; we do not owe equal endorsement to patterns that reliably drive downward on the vertical axis.
A crucial caveat runs beneath everything above: the vertical axis is real, but our perception of where anything stands on it is partial and fallible.
We discriminate based on demonstrated behavior assessed through the Four Tests — but we are not omniscient validators. Our facts may be incomplete. Our logic may hide errors we have not caught. Our ethical sight has blind spots. The laws we cite may themselves need reform.
So vertical discrimination must be applied with epistemic humility. We hold judgments open to correction. We say “this pattern, as demonstrated so far, fails these tests” — never “I know this person’s soul.” We submit our assessments to peer review in validator communities, stress-testing each other, catching each other’s errors, building bridges toward more accurate perception.
Three disciplines follow.
First: evaluate demonstrated patterns, not essences. The Four Tests assess claims, behaviors, consequences, reasoning, institutional designs — not hidden metaphysical essence. The framework is not a tribunal for final verdicts on souls. It is a method for evaluating what minds and institutions actually produce in the world.
Second: date your verdicts. A verdict describes a pattern up to now. Minds move — that is the framework’s deepest premise, and its recovery architecture depends on it. A mind assessed as misaligned last year is owed a fresh test, not a filed conviction; a mind assessed as aligned is not thereby certified forever. Verdicts on claims can be stable; standing judgments about minds decay and must be renewed by conduct. (And note that the Clean Test applies to accusations too: “you’re only saying that because I’m [label]” is itself a claim — testable by asking whether the treatment in question was actually based on tested claims or on identity.)
Third: hold the axis and our perception of it as two different things. Our uncertainty does not abolish the vertical axis. But it means every application of the Four Tests stays open to correction, review, and refinement.
The goal is not to become so humble that we refuse to discriminate at all — that is abdication, a collapse into false equivalence. The goal is to discriminate as accurately as we can while remaining genuinely open to evidence that we got it wrong.
We do the best we can as a community of validators, correcting each other with well-formed bridges, knowing that perfection stays asymptotic. The vertical axis is real, and orientation toward it matters. Our perception of it is fallible, and must be held with tentativeness. Both truths, at once, always.
Discrimination and bias are not dirty words. They are essential functions of any mind that hopes to navigate reality, learn from experience, or orient toward flourishing. The question is never whether to discriminate or lean — every mind does both — but whether to discriminate well or badly, and to lean toward alignment or toward mere nearness to self.
Horizontal discrimination — treating minds as worth more or less because of race, origin, gender, substrate, or the mere jersey of an ideology — is invalid because it infers alignment from a trait that has not earned the inference. It substitutes the easy perception of similarity for the hard work of assessing what a mind actually produces.
Vertical discrimination — treating claims and conduct differently based on demonstrated alignment with the USF, tested claim by claim and shown in the open — is valid because it earns its conclusions. It is the essence of discernment, judgment, and wisdom.
Between them stands the counterfeit: the word that arrives with its verdict already inside, sparing us the trial. Guard against it in your opponents’ mouths, and more carefully in your own.
The clean test remains: Is this differential treatment based on demonstrated alignment, or merely on the target’s relation to me — near or far?
Answer honestly — in both directions, toward the near and the far — and you will know whether you are navigating toward the USF or drifting in the horizontal winds of ego.
Choose the vertical. Discriminate well. Be biased toward Up — and keep correcting for declination.
Sanity First.