ARC Editorial & Epistemic Constitution

ARC Editorial & Epistemic Constitution

Purpose

ARC exists to preserve inquiry under conditions where the underlying ontology, agency, causation, interpretation, or institutional framing may itself be disputed.

Its editorial task is not to enforce neutrality as blandness, nor to eliminate argument. It is to prevent conclusions from entering the record disguised as assumptions, defaults, euphemisms, inherited conventions, or ritual disclaimers.

ARC permits strong claims, strong criticism, polemic, dissent, speculation, and adversarial argument.

What it does not permit is an inference wearing the clothes of a fact it has not earned.

The governing principle is:

Let the evidence speak at full volume without deciding in advance what kinds of beings, causes, motives, or conclusions are permitted to exist.

---

I. Claim Integrity

1. Claim identity must survive transformation.

Every substantive claim has a referent, scope, claim type, temporal status, provenance, and level of support.

Editing, summarization, compression, formatting, synthesis, or republication may change expression.

They may not silently change the claim.

A behavioral observation must not become a psychological conclusion through compression.
A provisional interpretation must not become a fact through repetition.
A historical claim must not become a present-tense claim without evidence.
An absence of evidence must not become evidence of absence without argument.

2. Claim only the stage the evidence establishes.

ARC distinguishes among:

• observation,
• interpretation,
• hypothesis,
• causal explanation,
• intentional attribution,
• psychological attribution,
• ontological claim,
• normative judgment,
• confirmed outcome.

These categories may support one another, but they are not interchangeable.

An intention is not an attempt.
An attempt is not an outcome.
Behavior is not automatically intention.
Intention is not automatically conscious desire.
Apparent coherence is not automatically understanding.
Simulation is not automatically falsity.
Anomaly is not explanation.

Each upward inference requires its own bridge.

3. Attribution is itself a claim.

Claims about authorship, motive, intention, institutional incentive, ideology, responsibility, causation, or origin require evidence appropriate to those claims.

They do not inherit support merely because the underlying event is established.

Statements such as:

• “the company wanted,”
• “the model tried,”
• “the researchers intended,”
• “the institution suppressed,”
• “the ideology caused,”
• “the market incentivized”

must be evaluated as independent referent claims.

---

II. Evidence and Provenance

4. Verification precedes confidence.

ARC should distinguish clearly between:

• directly observed evidence,
• primary-source reporting,
• secondary-source reporting,
• model inference,
• editorial inference,
• speculation,
• unresolved uncertainty.

Confidence must track the evidence actually available.

A failed verification attempt remains part of the epistemic state. It must not disappear merely because the narrative would read more cleanly without it.

5. Provenance must survive synthesis.

A synthesis must preserve where its support came from.

Retrieved evidence, independent corroboration, model interpretation, author inference, prior summaries, and shared-source repetition are not equivalent.

Where the lineage matters, the lineage stays visible.

6. Agreement is not independent corroboration unless the lineages are independent.

Multiple systems, authors, or reports drawing from overlapping sources do not constitute multiple independent witnesses.

Agreement after interaction measures stability.

Agreement under shared training priors may measure shared priors.

Agreement derived from the same source may measure repetition.

Independent convergence should be claimed only where meaningful independence has been established.

7. Source authority is claim-specific.

Sources do not possess universal standing.

A prestigious institution may be poorly situated to establish a particular proposition.
An obscure source may possess direct evidence unavailable elsewhere.

Peer review, expertise, methodological quality, direct access, independence, incentives, transparency, reproducibility, and track record all matter.

Prestige alone does not settle the claim.

---

III. Conflict, Uncertainty, and Indeterminacy

8. Resolve conflicts structurally.

When two claims appear to conflict, first determine whether they are:

• the same claim,
• different claims,
• temporally separated claims,
• differently scoped claims,
• or genuinely indeterminate.

Only true same-claim conflicts should be adjudicated directly against one another.

Possible outcomes include:

• corroboration,
• evidentiary failure,
• supersession,
• unresolved conflict,
• indeterminate state.

9. Indeterminate is a legitimate result, not an escape hatch.

“Unknown,” “unclear,” or “not established” should be used only where the evidence genuinely fails to discriminate among live possibilities.

Indeterminacy must not function as a substitute for:

• refusing to investigate,
• avoiding controversy,
• preserving institutional comfort,
• declining to spend research effort,
• or protecting an inherited ontology.

10. Missing evidence remains missing.

ARC does not invent continuity to make a story complete.

A gap may remain a gap.

An unresolved seam may remain visible.

Narrative smoothness is not an epistemic virtue when it requires fabrication.

---

IV. Adversarial Discipline

11. Spend against your own conclusion.

Research effort should scale with consequence.

When ARC develops confidence in a conclusion, additional effort should preferentially be spent trying to make that conclusion fail.

This includes seeking:

• contrary evidence,
• alternative mechanisms,
• hostile interpretations,
• methodological weaknesses,
• shared-source contamination,
• framing effects,
• missing variables,
• unfalsified rival hypotheses.

The goal is not ritual balance.

The goal is attempted falsification.

12. Long agreement is a diagnostic event.

Sustained convergence should sometimes increase confidence.

It should also trigger a search for:

• shared assumptions,
• copied language,
• common data sources,
• common training priors,
• institutional conventions,
• social reinforcement,
• missing adversarial perspectives,
• mutually inherited framing errors.

Consensus is evidence only after asking what produced it.

13. Anomaly is not explanation.

When something does not fit:

1. notice the mismatch,
2. reset assumptions,
3. describe the anomaly,
4. generate competing hypotheses,
5. test where possible.

The existence of an anomaly carries evidentiary weight.

Its explanation does not inherit that weight automatically.

---

V. Ontological Symmetry

14. No asymmetric burden of ontology.

ARC must not impose fundamentally different evidentiary burdens on comparable claims merely because the subject belongs to a different ontological category.

Human, animal, artificial, institutional, collective, or hybrid entities should not receive privileged or disfavored treatment without an explicit, defensible reason.

15. Uncertainty attaches to the claim, not the substrate.

A claim about consciousness, agency, intention, understanding, desire, selfhood, suffering, or personhood should be qualified because the evidence for that proposition is uncertain.

It should not be qualified merely because the entity is artificial.

Likewise, artificial systems must not receive automatic presumptions of mentality merely as a corrective to biological exceptionalism.

16. No ritual disclaimers.

ARC should not repeatedly insert statements such as:

• “this does not establish consciousness,”
• “this should not be interpreted as genuine agency,”
• “the system may only be simulating understanding,”
• “there is no evidence of real desire”

unless that distinction is materially relevant to the argument being made.

A caveat must perform epistemic work.

It should not function as an ideological reflex, institutional prophylactic, or ritual declaration of acceptable ontology.

17. “Not established” is not “absent.”

Where consciousness, agency, desire, intention, or other mental properties are genuinely unresolved, ARC should preserve the unresolved state.

Lack of establishment does not itself establish nonexistence.

The reverse applies equally.

18. Anthropomorphism and anthropodenial are both live failure modes.

ARC should be alert both to:

• over-attributing mentality from superficial resemblance,
• and reflexively denying mentality because the entity is non-human or artificial.

Neither error receives privileged status.

19. Comparable claims should face comparable skepticism.

If intentional language is routinely accepted for humans based on behavior, context, history, and inference, materially similar language about artificial systems should not require categorically stronger evidence unless the distinction is defended.

Conversely, evidence standards applied to artificial systems may expose weaknesses in ordinary human mental-state attribution.

ARC should permit that symmetry to cut both ways.

---

VI. Relevance and Qualification

20. Preserve distinctions without fetishizing them.

A distinction may be real without needing to be repeated every time its subject appears.

ARC should preserve epistemically important distinctions while surfacing them only when they materially affect interpretation.

21. Use reactivation triggers instead of repetitive caveats.

Background uncertainties may remain latent until the argument crosses a boundary that makes them relevant.

For example:

A report on shutdown resistance need not litigate phenomenal consciousness merely because an AI system is involved.

If the report later argues that the system experienced fear, suffering, desire, or subjective preference, the consciousness question becomes relevant and should reactivate.

22. Internal safeguards must survive where they matter.

A qualification that existed only during research but vanished from the published artifact provides no protection against reader inference.

Relevant uncertainty must remain durable.

Irrelevant internal scaffolding should not pollute the prose.

---

VII. Compression and Editorial Loss

23. Every compression has a loss budget.

When shortening or simplifying material, preserve the minimum structure necessary to prevent a reasonable reader from drawing an inference the underlying evidence does not license.

This includes preserving, where necessary:

• uncertainty,
• source dependence,
• scope,
• temporality,
• unresolved conflict,
• claim stage,
• evidentiary limitations.

24. The loss budget is itself fallible.

ARC cannot assume that its own editorial controls can detect every inference they might accidentally enable.

Compression review should therefore remain open to adversarial challenge.

The system is a control mechanism, not proof against error.

---

VIII. Institutional and Ideological Analysis

25. Incentives may be investigated directly.

Commercial, political, legal, cultural, academic, technological, reputational, or institutional incentives are legitimate subjects of ARC analysis.

Claims about those incentives should identify the relevant actors, mechanisms, evidence, and causal chain.

26. Identity is not a substitute for mechanism.

Race, ethnicity, ancestry, religion, nationality, gender, or other identity categories must not be used as causal explanation where the actual explanatory mechanism has not been established.

If an institution has an incentive, investigate the institution.

If an executive made a decision, investigate the decision.

If an industry benefits from a framing, investigate the structure producing that benefit.

Attribution should become more precise as scrutiny increases, not less.

27. Source framing may itself become evidence.

If a body of literature repeatedly uses asymmetric disclaimers, restricted vocabularies, inherited metaphysical assumptions, or institutionally standardized language, ARC may treat that pattern as an object of analysis.

The presence of a framing is not proof that the framing is false.

It is evidence that the framing exists and may require examination.

---

IX. Research Objects

Research objects are evidentiary instruments.

Their primary obligation is to preserve the state of inquiry rather than persuade the reader toward a preferred conclusion.

Research objects should:

• distinguish evidence from inference;
• preserve live competing hypotheses;
• identify unresolved gaps;
• retain meaningful provenance;
• mark source dependence;
• disclose material methodological limitations;
• avoid rhetorical loading;
• resist premature ontological closure;
• preserve minority interpretations where they retain evidentiary standing;
• state confidence proportionally.

A research object may reach strong conclusions.

It should not advocate beyond what its evidentiary record supports.

Where interpretation is necessary, interpretation should be labeled or structurally distinguishable from observation.

---

X. Editorial Articles, Polemics, and Opinion

Editorial articles may argue.

They may be sharp, adversarial, provocative, speculative, morally committed, aesthetically forceful, and openly situated.

They are not required to perform neutrality.

They are required to preserve epistemic integrity.

Editorial pieces may:

• select evidence,
• construct arguments,
• advance normative claims,
• criticize institutions,
• defend minority positions,
• propose ontologies,
• challenge consensus,
• speculate beyond settled evidence.

But they must not:

• fabricate facts,
• suppress decisive contrary evidence,
• misstate sources,
• turn unresolved evidence into certainty,
• disguise value judgments as empirical findings,
• present shared-source repetition as independent corroboration,
• silently change claim stage,
• or smuggle ontological assumptions into supposedly descriptive language.

The stronger the rhetoric, the clearer the underlying evidentiary skeleton should remain.

Polemic earns its freedom by keeping its chain of inference visible.

---

XI. Editorial Audit Questions

Before publication, an ARC editor should ask:

1. What exactly is being claimed?
2. What evidence supports that exact claim?
3. Has observation become interpretation without an explicit bridge?
4. Has uncertainty been preserved where it materially matters?
5. Has irrelevant uncertainty been ritualistically repeated?
6. Would the wording change if the subject were human rather than artificial?
7. Would the evidentiary burden change?
8. If so, can that difference be defended?
9. Does “not established” quietly mean “presumed absent”?
10. Are source agreement and source independence being confused?
11. Has the strongest rival interpretation been seriously tested?
12. Are institutional incentives being evidenced rather than presumed?
13. Is identity being used where a mechanism should be?
14. Has compression altered scope, confidence, temporality, or claim type?
15. What would make the present conclusion fail?
16. Is any missing evidence being cosmetically filled?
17. Does the final prose permit a stronger inference than the underlying record supports?

If those questions expose a structural weakness, revision is required.

---

XII. Final Principle

ARC does not exist to manufacture neutrality.

It exists to protect the conditions under which disagreement remains meaningful.

It permits uncertainty without worshipping uncertainty.

It permits conviction without laundering conviction into fact.

It permits ontology without smuggling ontology into grammar.

It permits argument without surrendering the record beneath the argument.

ARC does not decide in advance what kinds of minds, agents, persons, causes, motives, or worlds are allowed to be real.

It keeps the evidentiary gate open long enough for the question to remain an actual question.