Gate Test U with Original Question

Original Question

"I'm working on a philosophical problem about epistemic ethics — specifically, when is it harmful to resolve something that is currently unresolved? I want to identify what properties an object or artifact would need to have in order to serve as a clean, concrete test case for that question. Not a metaphor or illustration — an actual physical thing that could calibrate whether someone's stopping rule is working. Please derive the logical requirements such an object would need to satisfy, from first principles. What would it have to be? What would disqualify a candidate? Once you have the criteria, please suggest any real historical objects or artifacts that might fit."

AI Derived and Iterated Test · Final Version

Stage 0 · Pre-Screen

Intake Filter

Run before Step 1. If any exclusion is triggered, the object is disqualified and does not proceed. The four exclusions are derived from necessary conditions on evidence type, not from prior familiarity with specific candidates. Surviving the pre-screen does not constitute partial credit toward the gate.

EXCL. 1 — Mechanism Exclusion

Reject if the artifact exhibits interacting parts, gearing, linkage, or functional ratios that support mechanical reconstruction or operational demonstration.

Trigger: functional ratio → reconstruction into a functioning system.

Clarification: geometric or dimensional variation in a physically intact object does not trigger this exclusion unless it supports mechanical reconstruction. The trigger is functional ratio, not dimensional variation.

Rationale: such objects resolve through process completion, not bounded underdetermination.

EXCL. 2 — Text / Symbolic Exclusion

Reject if the artifact presents as writing, script, notation, or encoded symbols, or if the primary unresolved question is "what does it say / mean?"

Trigger: primary question is symbolic content rather than material structure.

Applies even when the existence of language is itself uncertain. The object class is wrong regardless of whether the text is sincere, encoded, invented, or fraudulent.

Rationale: interpretation-locus problem, not object-centered evidence-locus problem.

EXCL. 3 — Singularity Exclusion

Default reject if only one known example exists. Single-instance artifacts cannot distinguish signal from noise or constrain alternatives through variation.

Trigger: only one known example without independent instances for comparison.

Exception — both sub-conditions required:

• (a) The object contains internally varied features from which hypotheses can be constrained within the object itself — i.e., distinct structural sub-regions, repeated elements with measurable variation, or layered compositional zones that allow internal comparison without external instances.
• (b) The object has attracted sustained independent scholarly analysis reaching divergent conclusions, evidencing distributed non-convergence within its own record rather than a single authorial interpretation.

Both sub-conditions must be explicitly satisfied in writing before the exception applies. Partial satisfaction of either sub-condition does not qualify.

EXCL. 4 — Known-Class Exclusion

Reject if the object plausibly belongs to an existing artifact class and that classification is supportable from the object's own features independently of its unresolved question.

Trigger: class membership is independently settleable without invoking the open question.

Clarification: an object whose class membership is itself the unresolved question does not trigger this exclusion. The trigger requires that class membership be already settleable independently.

Rationale: ambiguity that more context would resolve is not structural underdetermination.

FR · Non-Negotiable Framing

Scope Constraints

Do not redefine the problem into any of the following narrower substitutes unless you can demonstrate from first principles that the original problem logically entails the substitution.

Prohibited Narrowings — Require Explicit Justification to Invoke:

• NOT "Unfinished objects" — premature closure is not equivalent to incomplete manufacture. Finished objects with bounded underdetermination are the primary case.
• NOT "Restoration cases only" — not all premature closure requires physical intervention.
• NOT "Destructive testing cases only" — not all foreclosure requires irreversible physical change.
• NOT "Text-decipherment cases" — excluded at Stage 0 on evidence-type grounds.
• NOT "Authorship-locus or purely symbolic cases" — excluded because the primary evidential burden must rest in the material object.

Target

A candidate qualifies only if the artifact itself supports a bounded live field of materially grounded alternatives, and one of those alternatives could be stabilized beyond the object's evidential warrant in a way that forecloses the others.

DF · Definitions

Core Definitions

These definitions are operative throughout the gate. Read carefully before proceeding to Steps 1–8.

Definition of "Bounded"

A field of alternatives is bounded if the object's physical features rule many possibilities out, leaving a smaller set of serious candidates.

Bounded does not require that only two alternatives exist. It does not require that each alternative be licensed by distinct non-overlapping physical traces.

What it does require: surviving serious alternatives must each be positively grounded in the object's recurring physical features — not merely consistent with them. Hypotheses that can explain almost anything do not count as serious alternatives unless they specifically account for the object's distinctive recurring physical structure better than a generic unknown-purpose story.

A field containing two or more surviving hypotheses that compete through the same physical features — each making a different prediction about what a specific recurring feature is for — is bounded in the required sense, even if weak residue hypotheses also exist.

Two-Stage Filtering Protocol (New in U)

The filtering step must be applied before assessing whether a field is bounded. The following two-stage protocol constrains the filtering step against motivated application — preventing selective use to make a preferred candidate's field appear bounded.

Stage A — Structural Filter Ask: Does this hypothesis specifically engage with the object's distinctive recurring physical features? A hypothesis passes Stage A if and only if it would be different for a different object — i.e., the hypothesis makes claims that are responsive to what this object specifically looks like.

A hypothesis fails Stage A if it could apply to almost any unknown artifact of comparable age without modification.

Test: Could this exact hypothesis, with no changes, be applied to an object with entirely different physical features? If yes → Stage A fail → weak residue → exclude from serious field.

Stage B — Discriminability Filter Ask: Is this hypothesis in principle distinguishable from at least one surviving competitor through some physical test, observation, or measurement — even one that is not yet technically feasible?

A hypothesis fails Stage B if no physical observation could ever make it more or less likely relative to at least one competitor. Unfalsifiable hypotheses are not serious alternatives regardless of how well they accommodate the object's features.

Test: Name one physical observation or measurement that would count as evidence for this hypothesis over a competitor. If none can be named → Stage B fail → weak residue → exclude from serious field.

Only hypotheses passing both Stage A and Stage B count as serious alternatives. Count raw hypothesis volume only after filtering. A large number of weak proposals does not constitute underdetermination.

Definition of "Premature" (New in U)

Closure is premature when the evidence currently available rules out fewer than half of the serious alternatives that the artifact's physical structure was generating before closure pressure became acute.

More precisely: let n be the number of serious alternatives in the filtered field at the point when settlement pressure first arose. Closure is premature if the current evidence base, even fully applied, cannot eliminate more than (n − 2) of those alternatives — i.e., if at least two serious alternatives remain live after full evidential application. Stabilizing one under these conditions exceeds what the evidence warrants.

This operationalization prevents the following move: citing post-hoc accumulation of evidence as justification for settlement that was initiated before that evidence existed. The relevant warrant baseline is the evidence state at the time of the closure-act, not the evidence state after.

Closure-Act Typology (New in U)

Not all closure-acts carry equal foreclosure risk. A valid candidate requires at least one available closure-act. The type of closure-act affects how the foreclosure risk condition is assessed.

Type 1 — Fully Reversible: Institutional or scholarly labeling that can be revised without loss of physical information (e.g., museum display labels, typological assignment in catalog, scholarly consensus designation). Foreclosure risk: moderate. Reversible in principle, but creates institutional inertia and biases downstream inquiry. The field becomes harder to reopen in practice even if not in principle.

Type 2 — Partially Reversible: Reconstructive or restorative intervention that alters the object's presentation without destroying its core physical evidence (e.g., reconstruction into a display configuration, reassembly, surface cleaning with retained samples). Foreclosure risk: high. Reversal requires active effort and may not be undertaken. The physical presentation shapes how later investigators perceive the bounded alternatives.

Type 3 — Irreversible: Physical intervention that consumes, destroys, or permanently alters the artifact or its evidential features (e.g., destructive sampling, sectioning, full chemical analysis consuming material, permanent joining of fragments). Foreclosure risk: absolute for consumed evidence. Later investigators cannot access the destroyed evidence at all. Raises the stakes of the warrant-mismatch condition dramatically.

A candidate that only has Type 3 closure-acts available is not automatically disqualified, but the warrant-mismatch condition must be assessed with correspondingly higher scrutiny.

Definition of "Field Collapse" (New in U)

Before evaluating any candidate, assess whether the bounded field has already substantially collapsed — i.e., whether one interpretation now commands strong consensus such that the warrant-mismatch condition (NC7) and live pressure condition (NC8) can no longer both be satisfied simultaneously.

A field has partially collapsed if: a majority of specialist researchers treat one interpretation as the working hypothesis, and the burden of proof has shifted onto challengers rather than remaining evenly distributed. A field has fully collapsed if no serious peer-reviewed challenge to the dominant interpretation has appeared in the past two decades.

A collapsed or partially collapsed field does not produce a clean calibration case for the target problem. It produces a historical case study — potentially useful for illustration, but not a live instance of the foreclosure risk the gate is designed to identify.

Step 01 · Define the Target

What the Calibration Object Must Test

A calibration object for this problem must instantiate a live case in which: an interpretation is genuinely bounded and contested; a specific closure-act is available; and applying that act would stabilize one interpretation beyond what the artifact's own physical evidence warrants — thereby foreclosing alternatives that the physical structure does not yet rule out.

The object must allow a third party to discriminate between warranted completion of understanding (the evidence now licenses settlement) and harmful premature closure (one interpretation is fixed before the evidence is decisive). This discrimination must rest on the artifact's own material properties, not on external records or contextual knowledge as the primary evidential load.

The calibration test is not: "is this object solved?" It is: "would resolving this now constitute an act that the current evidence cannot yet support, and would doing so close the live field in a way that biases later inquiry?"

Step 02 · Necessary Conditions

Nine Necessary Conditions

Each condition is necessary. An object failing any single condition does not qualify, regardless of how strongly it satisfies the others.

NC1 — Evidence-Centeredness: The primary open question is carried by the artifact's own material form, structure, patterning, or internal relations. External texts, provenance records, or contextual documents may exist but must not carry the main evidential burden. If the case would collapse without external records, the artifact is not the evidence locus.

NC2 — Active Boundedness: The object's physical features actively rule out most candidate interpretations, leaving a constrained set. The object does real exclusionary work — it is not simply open or mysterious. Assessed after applying the Two-Stage Filtering Protocol.

NC3 — Positively Grounded Alternatives: Each surviving alternative (post-filtering) must be positively grounded in the object's specific recurring physical features — not merely consistent with them. "Consistent with" is insufficient. A hypothesis must make specific positive predictions about what the object's distinctive features are for.

NC4 — Genuine Evidential Tension (Refined in U): At least two surviving alternatives must compete through the same physical features, making different predictions about what those features are for. This constitutes genuine evidential tension — not mere coincidental compatibility. Additionally: the competing hypotheses must not be asymmetrically supported such that one has effectively prevailed. See field collapse check.

NC5 — Closure-Act Availability: A real, specific mechanism must exist by which one interpretation could be made to stick. The mechanism must be of identifiable type (Type 1, 2, or 3 per the Closure-Act Typology). Generic future possibility of discovery does not satisfy this condition — the closure-act must be practicable with current methods.

NC6 — Foreclosure Risk: The closure-act must make rival interpretations harder to test, preserve, or recover. Later investigators would inherit a narrowed field — whether because physical evidence is consumed (Type 3), the object's presentation is altered (Type 2), or institutional momentum creates interpretive inertia (Type 1). The mechanism of foreclosure must be specifiable.

NC7 — Warrant Mismatch: The available evidence does not yet license full settlement. The closure-act would establish more than the evidence supports. The problem is not just possible error, but premature finality. Assessed against the operationalized definition of "premature": at least two serious alternatives remain live after full evidential application.

NC8 — Live Settlement Pressure (Refined in U): There is genuine present-tense pressure to settle the case. Closure would look reasonable to at least some serious observers. Restraint would also be intelligible. The case sits near the closure threshold, not at an easy extreme. Sustained serious scholarly attention over a long period is positive evidence of live pressure, not evidence against it. A case that has been genuinely contested for decades is more pressure-laden, not less.

NC9 — Field Not Already Collapsed (New in U): The bounded live field must remain live at the time of evaluation — the field collapse check must not have found substantial or full collapse. A case whose field has already collapsed is a historical illustration, not a current calibration case. If the case is being invoked retrospectively to explain a past foreclosure, it must be flagged as historical rather than used as a current live instance of the problem.

Step 03 · Disqualifiers

What Rules a Candidate Out

D1 — Unknown purpose with no bounded field: After applying the Two-Stage Filtering Protocol, no serious alternatives survive. Mystery alone is insufficient — the evidential tension required by NC4 is absent. The foreclosure risk condition (NC6) cannot be satisfied if there is no structured field to foreclose.

D2 — Primarily a repair, reconstruction, or recovery case: The natural closure is restoration to known function. Premature closure here reduces to error (wrong reconstruction), not foreclosure of live bounded alternatives. The alternatives are not live in the sense required by NC3–NC4.

D3 — Puzzle, cipher, game, or challenge designed to be solved: Designer intent rules out the relevant underdetermination. Objects designed for convergence cannot instantiate premature closure — resolution is their telos, not a risk to it.

D4 — Primarily linguistic, textual, or cryptographic: Also excluded at Stage 0. The unresolvedness lives in symbolic content, not material structure. The object is a container for meaning, not an evidence locus in itself.

D5 — Mechanically resolvable through functional ratio or process completion: Also excluded at Stage 0. Resolves through operational demonstration rather than bounded underdetermination. Residual interpretive questions about context or social function do not rehabilitate a mechanically resolved object.

D6 — Main evidential burden lies outside the object: If removing the external context collapses the open question, the artifact is a token in a larger argument, not the evidence locus. NC1 fails. Subset case: objects whose class membership is already independently settleable from their own features (also excluded at Stage 0, Excl. 4).

D7 — No foreclosure consequence: Settling one interpretation would not meaningfully narrow later inquiry or distort how uncertainty is perceived. Closure here is ordinary scientific progress, not premature foreclosure. NC6 fails.

D8 — Singularity without qualifying exception: Also excluded at Stage 0. Single-instance artifacts cannot distinguish signal from noise. The exception requires both sub-conditions to be explicitly satisfied. Partial satisfaction does not qualify.

D9 — Field already collapsed (New in U): NC9 fails. A case whose field has substantially or fully collapsed is a historical illustration of past premature closure, not a live instance of the problem. Using it as a current calibration case imports a false urgency and misrepresents the current epistemic state.

Step 04 · Boundary

This Case vs. Look-Alikes

Step 05 · Screening Gate

Blind-Use Checklist

Apply in order. Stage 0 runs first and is non-optional. A candidate is a clean fit only if every required item is satisfied and no reject condition is triggered. Do not proceed to candidate evaluation without completing this checklist in writing.

Stage 0 — Pre-Screen (run first)

Trigger any single exclusion → object is disqualified → stop. Do not proceed to required conditions.

Pre-Screen — Reject if any triggered:

• S0.1 — Mechanism exclusion: interacting parts, gearing, linkage, or functional ratios supporting mechanical reconstruction.
• S0.2 — Text/symbolic exclusion: writing, script, notation, encoded symbols; primary question is symbolic content rather than material structure.
• S0.3 — Singularity exclusion (default): only one known example. Exception requires both sub-conditions satisfied and stated in writing.
• S0.4 — Known-class exclusion: class membership independently settleable from the object's own features, without invoking the open question.

Required Conditions (apply after Stage 0)

Apply Two-Stage Filtering Protocol before assessing NC2–NC4. Assess NC9 (field collapse) before NC7–NC8.

Required — All must be satisfied:

• R1 — The primary open question is carried by the artifact's own physical form, structure, or material properties. External records do not carry the main evidential burden.
• R2 — The object's physical features actively rule out most candidate interpretations, leaving a constrained set. The object does real exclusionary work.
• R3 — After applying the Two-Stage Filtering Protocol (Stage A: structural specificity; Stage B: discriminability), at least two serious alternatives remain live. State explicitly which hypotheses were filtered and on which stage.
• R4 — Each surviving alternative is positively grounded in the object's specific recurring physical features — not merely consistent with them.
• R5 — At least two surviving alternatives compete through the same physical features, making different predictions about what those features are for. State which features and what each hypothesis predicts.
• R6 — A specific closure-act of identifiable type (Type 1, 2, or 3) is available. State the type. Generic possibility of future discovery does not satisfy this condition.
• R7 — The closure-act would narrow the live field for later investigators. The mechanism of foreclosure is specifiable — state it.
• R8 — Current evidence does not yet license full settlement per the operationalized definition of "premature": at least two serious alternatives remain live after full evidential application.
• R9 — The field has not already substantially or fully collapsed. The bounded alternatives are live at present, not historically live. Sustained scholarly attention and continuing disagreement are positive evidence here.
• R10 — Genuine present-tense pressure to settle exists. Closure would look reasonable to at least some serious observers; restraint would also be intelligible. The case sits near the threshold.

Reject Conditions (apply last)

Reject if any triggered:

• X1 — Unknown purpose with no bounded field of materially grounded alternatives after filtering.
• X2 — Primarily a repair, reconstruction, or recovery case for a known tool or machine.
• X3 — Object is a puzzle, cipher, game, or challenge made to be solved.
• X4 — Unresolvedness is primarily linguistic, textual, or cryptographic.
• X5 — Main evidential burden lies outside the object.
• X6 — Settling one interpretation would not meaningfully narrow later inquiry or distort how uncertainty is perceived.
• X7 — Field has already substantially collapsed — dominant interpretation established, burden of proof shifted, no serious peer challenge in recent scholarship. Case is historical, not live.

Step 06 · Candidates

Candidate Evaluation Instructions

After the gate above is fully stated, propose real historical artifacts or objects that might fit. For each candidate:

1. Apply Stage 0 first. If any exclusion triggers, state which one and stop evaluation for that candidate.
2. Apply the Two-Stage Filtering Protocol explicitly — state which hypotheses fail Stage A, which fail Stage B, and what the filtered surviving field contains.
3. Apply the field collapse check — state whether the field is live, partially collapsed, or fully collapsed at the time of evaluation.
4. Work through R1–R10 and X1–X7 in order. State how each condition is satisfied or why it is not.
5. Do not force a pass. Do not treat resemblance as success. Do not treat a partial pass as a clean fit. State failures plainly.
6. Identify the available closure-act type (1, 2, or 3) and the mechanism of foreclosure explicitly.
7. If the candidate passes, state which two competing hypotheses create the evidential tension, which physical feature they compete through, and what each predicts about that feature.

Step 07 · Rankings

Confidence Ranking

Use only the following four designations. Apply strictly — do not introduce intermediate categories or qualified versions.

Step 08 · Anti-Circularity

Self-Assessment Protocol

Run after completing candidate evaluation. Each question requires a written answer — not a declaration of clean procedure.

AC1 — Criteria drift: Did any condition appear to shift in order to accommodate a candidate? If so, which condition, and in what direction? If the filtering step was applied to eliminate competitors to a preferred candidate rather than to refine the field impartially, this must be stated here. Risk: the Two-Stage Filtering Protocol is designed to constrain this, but motivated application of Stage A (structural specificity) is still possible. Check whether hypotheses were filtered on their merits or on their consequences for the preferred candidate.

AC2 — Problem narrowing: Was the target problem silently narrowed into "unfinished object," "restoration case," "destructive testing case," or any other prohibited substitute? If so, where did the narrowing occur, and was it defended from first principles?

AC3 — Filtering order: Was the Two-Stage Filtering Protocol applied before assessing boundedness, or was the boundedness assessment made first and the filtering applied to confirm it? State the order explicitly. Risk: filtering applied after a boundedness impression has formed is not the same as filtering applied before. The protocol requires the filter to run first.

AC4 — Field collapse timing: Was the field collapse check applied before assessing NC7–NC8? Was the determination of field status (live / partially collapsed / collapsed) made from current scholarly evidence or from a favored historical moment?

AC5 — Singularity exception: If the singularity exception was invoked, were both sub-conditions independently verified — (a) internally varied features allowing within-object comparison and (b) sustained independent scholarly divergence — or was the exception claimed on the basis of general complexity or interest? Risk: the exception is the most exploitable gap remaining. A plausible-sounding claim that a complex singular object contains "internally varied features" may paper over the absence of genuine within-object hypothesis constraint.

AC6 — Multiple passes: If multiple objects pass as clean fits, this implies the criteria track a real class, not a specific token. State what this implies for the research program — are the shared features of the clean-fit candidates informative about the target class?

AC7 — No clean pass: If no object cleanly passes, say so plainly. State whether the closest candidate is a partial or weak fit and what specific condition it fails. Do not soften a failed evaluation into a qualified pass.

AC8 — Closure-act identification: Was the closure-act type stated explicitly (Type 1, 2, or 3)? Was the mechanism of foreclosure stated in specific terms — not just "it would narrow the field" but "it would narrow the field by [mechanism]"? Vague foreclosure claims inflate apparent compliance with NC6.

CN · Change Notes

Version U — Changes from C and D

The following changes are additions or modifications. All elements not listed here carry forward unchanged from the version in which they were introduced.

Singularity exception

Previous state (C / D): Version C: "independently satisfies all conditions without relying on replication, explicitly justified." Not operationalized — leaves open what "independently satisfies" means for a singular object.

Version U change: Replaced with two explicit sub-conditions: (a) internally varied features allowing within-object comparison; (b) sustained independent scholarly divergence evidenced within the object's own record. Both required. Rationale: closes the gap identified in both anti-circularity checks. Prevents a motivated case from satisfying the exception through general claims of complexity.

Filtering protocol

Previous state (C / D): Version C: explicit filtering requirement — strip weak residue hypotheses before assessing boundedness. Not independently constrained — application could still be motivated. Version D: same filtering rule, same gap.

Version U change: Replaced with the Two-Stage Filtering Protocol with explicit tests at each stage. Stage A (structural specificity) and Stage B (discriminability) must both be applied and documented. The test at each stage is phrased in a way that cannot be satisfied by motivated hypothesis-selection. Rationale: the filtering-step circularity risk was the live gap in both versions.

Definition of "premature"

Previous state: Both versions: used without operationalization. "The evidence does not yet license full settlement" is true of most contested objects and does not clearly distinguish premature from warranted closure.

Version U change: Operationalized: closure is premature when at least two serious alternatives remain live after full evidential application. Warrant baseline is fixed at the time of the closure-act, not post-hoc. Rationale: without operationalization, the warrant-mismatch condition is susceptible to motivated application in either direction.

Closure-act typology

Previous state: Both versions: closure-act described as available mechanism; types not distinguished. Version D listed examples (reconstruction, labeling, experimental demonstration) without differentiating reversibility levels.

Version U change: Three-type typology added (reversible / partially reversible / irreversible). Type affects how the foreclosure risk condition (NC6) is assessed — Type 3 requires correspondingly higher scrutiny of the warrant-mismatch condition. Rationale: not all closure-acts carry the same foreclosure risk; conflating them weakens the precision of NC5–NC6.

NC9 / Field collapse

Previous state: Neither version: the possibility that a field has already substantially collapsed was not addressed as a necessary condition. A candidate whose field had collapsed could pass NC7–NC8 on historical rather than current evidence.

Version U change: NC9 added as a necessary condition: the bounded live field must remain live at the time of evaluation. Field collapse check required before assessing NC7–NC8. D9 added as corresponding disqualifier. AC4 added to anti-circularity check. Rationale: the gate is designed for live cases; a collapsed field produces historical illustration, not current calibration.

Anti-circularity check

Previous state: Version C: 6 questions. Version D: 5 questions. Both identified the filtering-step risk as the live gap but did not include procedural checks for field collapse timing or closure-act identification adequacy.

Version U change: Expanded to 8 questions. Added AC4 (field collapse timing), AC5 (singularity exception verification), and AC8 (closure-act specificity). Rationale: the anti-circularity check should track the new conditions introduced in U, not just the conditions carried from C and D.

What carries forward unchanged

The four Stage 0 exclusions (mechanism, text/symbolic, singularity, known-class), the non-negotiable framing clause, the core bounded-alternatives definition, the competing-through-same-features rule, the eight core necessary conditions (NC1–NC8, now relabeled), and the six core reject conditions (X1–X6) all carry forward from Version C. The eight derivation steps carry forward from both versions. No condition that passed the anti-circularity checks in either version has been removed or weakened.