Due Process for Ethics

An ethical judgment can become more trustworthy through a process that represents the case faithfully, examines the evidence, hears affected people, gives reasons and remains open to correction. Authority to impose a consequence requires a separately established mandate, grounds for the particular action and a defined scope. Calling an intervention ethical places its own conduct inside that examination.

The ACCM Deep Ethics Project develops this relationship across human inquiry, institutions and artificial intelligence. Its orientation is omnidirectional, high-signal, deep-ethical, mutually dignifying, corrigible, sense-making and process-oriented. Procedural care helps people recover reality, contribute something useful and retain agency when consequential judgments are made about them.

Status: living conceptual and practical synthesis, commissioned by John Kuhles on 4 November 2026. The legal sources below provide identifiable comparison objects. The proposed ethical applications, teaching specimens and test conditions retain their own status. Legal duties and remedies depend on the jurisdiction, proceeding, actor and governing rule.

Reading routes: John’s connection · The innocent person · Legal relationships · Historical safeguards · Questions that stay distinct · Project instruments · 1888 in operation · Competing A.I. risks · Worked specimen · Remedies and stopping · Research connections · A usable case record · Proposed comparisons · Revision record.

Why John sees the connection

John began with a familiar Dutch newspaper story: a case can be affected by a vormfout, a procedural error. He connected that account to due process in the United States, then asked:

Why not apply the same logic to AI ethics washing mannerisms?

— John Kuhles 🦜🦋🌳, 4 November 2026

The connection concerns the authority exercised between an object and a consequential decision. In a legal setting, how evidence was obtained, how an allegation was framed, which opportunities existed to respond and whether the decision-maker acted within a mandate can affect what may legitimately follow. A desired outcome does not settle those questions.

An ethical evaluator faces a related responsibility. It selects evidence, interprets a person’s statement, applies categories, gives some considerations prominence and may recommend or impose a restriction. An A.I. can perform several of these operations in a single answer. Each transformation can change the case being assessed.

John recognizes a shared structural question: what must remain intact before the evaluator’s judgment earns authority over the person, claim or inquiry? The project’s recurring question follows the same transformation:

What happened to the object while we were processing it?

Recovering that path can reveal a warranted finding, a useful contribution, a correctable local mistake, an unsupported intervention or an unresolved case. The examination must be capable of reaching each result. Its purpose includes enabling a better next move, rather than keeping every inquiry permanently on trial.

The person called a criminal may be innocent

John added the clarification that governs this page’s treatment of the legal analogy:

By the way, sometimes an “assumed criminal” could be innocent and fall prey to due process failure/abuse!

— John Kuhles 🦜🦋🌳, 4 November 2026

The familiar story about someone escaping punishment on a technicality can already contain a verdict: it calls the person a criminal before establishing guilt in the case being discussed. A different story may be present. A person may have been misidentified, confronted with unreliable evidence, denied a meaningful response or judged through a distorted account. Procedural failure can help produce an unjust accusation or conviction, rather than merely interfere with punishment.

Article 6(2) of the European Convention on Human Rights protects the presumption of innocence for a person charged with a criminal offence. The Court’s presumption-of-innocence guide also distinguishes protections during proceedings from protections against later official treatment as guilty following acquittal or discontinuance. The label attached to the person cannot do the work of proving the case.

The ethical connection extends to classification outside court. Someone can become the subject of an allegation, suspicion, risk label or reputational judgment before the underlying object has been faithfully reconstructed. If the classification then governs what evidence is admitted, how their explanation is received and how much work they must perform to be heard, an initial mistake can organize the rest of the encounter.

Dignity and an opportunity to respond can be preserved while a claim remains unproven. Respect for a person does not require accepting their scientific hypothesis, factual account or proposed remedy. Conversely, disagreement with that contribution cannot by itself establish dishonesty, dangerousness or incompetence.

The affected field includes complainants, possible victims, accused people, witnesses, other exposed parties and those responsible for a decision. Procedural integrity helps examine the actual case among them. Their interests, vulnerability, access and decision rights can differ; faithful treatment makes those differences visible.

What the legal comparison contributes

Dutch procedural-error rules, U.S. constitutional due process and evidence-exclusion doctrines have related concerns and distinct legal operations. Keeping their boundaries visible makes the analogy more useful.

Legal source or doctrine	Relationship available for comparison	Consequence for the ethical inquiry
Dutch Article 359a framework	The Hoge Raad’s 1 December 2020 judgment considers the protected interest, seriousness of the procedural breach and resulting disadvantage. Available responses include recording the breach and, under the relevant conditions, sentence reduction, evidence exclusion or prosecution inadmissibility.	Identify what failed, what it affected and which response is warranted. A defect does not automatically require the most extensive remedy.
U.S. procedural due process	In Mathews v. Eldridge, the Court considered the affected private interest, risk of erroneous deprivation and value of additional safeguards, alongside governmental interests and procedural burdens.	Examine errors and the costs of protection together. Who can lose what? What would another safeguard actually improve?
Fourth Amendment evidence exclusion	Herring v. United States expressly distinguishes a constitutional violation from an automatic requirement to suppress evidence.	Keep the finding of a defect separate from selecting its consequence. Evidence exclusion is a particular legal doctrine, rather than a synonym for all due process.
Presumption of innocence	Article 6(2) protects a person charged with a criminal offence from being treated as proven guilty through the accusation alone.	Keep allegation, evidentiary assessment and finding distinguishable. Examine when a label has begun substituting for proof.

U.S. constitutional due process constrains governmental action in its applicable settings. The ethical proposal here concerns how evaluative power should be exercised across a wider field, including private A.I. services. The comparison does not automatically give a conversational exchange the legal status of a criminal proceeding.

The more general relationship is nevertheless practical: a factually accurate result and a legitimate use of that result require different examinations. Information can be true while its acquisition or use violates an applicable protection. A procedurally careful inquiry can also reach a mistaken conclusion that later evidence requires it to revise.

“Brought in Answer” — how safeguards change over time

The recurring question is what has to happen before a powerful actor may make an adverse decision effective. The documents below supply different answers about evidence, jurisdiction, participation and redress. Some belong to the English and American history of due process of law. The Dutch Plakkaat supplies a parallel comparison of judicial process and conditional authority.

From a restraint on royal action to an opportunity to answer

Source	What the text puts under examination
Magna Carta, 1215	The Articles of the Barons, article 29 give a draft predecessor to clause 39, which restricted specified actions against free men to lawful judgment of peers or the law of the land. Clause 38 addressed the grounds for opening a case: a bailiff’s accusation alone was insufficient. Clause 52 addressed restoration for certain royal dispossessions and identified who would hear a dispute over it. The clauses distinguish intake, judgment and repair.
Edward III’s 1354 statute	The traditional translation says that a man, whatever his estate or condition, may not suffer specified deprivations “without being brought in Answer by due Process of the Law.” The wording broadens the stated protection relative to free man and makes the timing of the answer visible. The Library of Congress account situates the phrase among several statutory restatements.
Edward Coke’s Second Institutes	In the passages reproduced in Hurtado v. California, Coke explained law of the land through later statutes and common-law forms, including presentment, indictment and original writ. These are more specific operations than the wording of clause 39 by itself supplies. How far a later use of the phrase imports those operations became a question for subsequent courts.

The English statute abolishing the Star Chamber in 1641, catalogued as the Habeas Corpus Act 1640, recites Magna Carta and several Edward III statutes, including the 1354 protection. Its preamble says the tribunal exceeded its authority and identifies ordinary justice as a route to remedy the matters it had heard. The law addresses both the loss of a tribunal’s jurisdiction and where legitimate work could continue.

The Plakkaat and the American constitutional documents

In the Plakkaat van Verlatinghe of 26 July 1581, the States General alleged that people were taken out of Brabant against its privileges, charged before someone without the proper jurisdiction and condemned without being fully heard. After stating the ruler’s duties, the alleged breaches and failed attempts at redress, they declared Philip II’s authority forfeited. The operative provisions discharged officials from their prior oaths, specified new oaths under the relevant jurisdiction and provided for continuity in certain offices. These are distinct questions about earlier acts, the continuing mandate and the administration of pending work.

The Dutch National Archives preserves Jan van Asseliers’s working draft and traces it to a resolution of 22 July. J.P.A. Coopmans’s comparison of the two declarations tracks their drafting and parallel grievances. Stephen E. Lucas argues for Dutch influence. Coopmans judged the evidence for direct borrowing by Jefferson inconclusive. The parallel can be inspected independently of that historical question.

American source	Change in the problem or the safeguard
Declaration of Independence, 1776	The Declaration justifies severing allegiance. It alleges that judges depended on the king for tenure and pay, that colonists were denied jury trials, and that troops were protected by a mock trial. The last two grievances make scrutiny visible in both directions: a safeguard can be withheld from one group and performed without accountability for another.
U.S. Constitution, 1787	The Constitution sets up a government with specific constraints. Article III protects federal judges’ tenure during good behaviour and their compensation in office. Article I bars federal and state bills of attainder, a form of legislative punishment without judicial trial. These provisions address who controls the reviewer and whether a rule-maker can bypass adjudication.
Bill of Rights, 1791	The Fifth Amendment establishes the federal due-process guarantee. The Sixth identifies practical ways an accused person can answer in a criminal prosecution, including notice, confrontation, defense witnesses and counsel. The due-process wording was already in Madison’s June 1789 proposal. Ryan C. Williams finds little recorded congressional debate about the phrase itself.
Fourteenth Amendment, 1868	During Reconstruction the amendment directly bound states and protected any person through its Due Process Clause. In the Bingham–Hotchkiss debate, Hotchkiss argued for a constitutional guarantee that a later Congress could not undo through ordinary legislation. The final amendment states the protection in Section 1 and the congressional enforcement power in Section 5. This makes the durability of a safeguard under changed leadership an explicit design question.
Hurtado v. California, 1884	The Supreme Court majority held that the Fourteenth Amendment did not require a state grand-jury indictment in the case before it. It read Coke’s specific forms as illustrations rather than an exhaustive list. The case makes a question portable: when the mechanism changes, which protection must the new procedure still deliver?

Answerability in both directions

Can the affected person answer the actual grounds while their response can still alter the move?
Can the authority answer for its evidence, rule, jurisdiction and consequences?
If authority changes hands, can the person still reach a remedy through an identifiable custodian?

These are proposed ethical tests. The historical sources supply comparison objects. In a flagging case, the person’s answer must reach the decision-maker with the relevant source. The authority’s answer must reach a reviewer who can examine what gave the flag consequences. The worked specimen below follows what happens when either route needs repair.

Four questions that remain distinct

Question	Object of examination	What answering it leaves open
What is supported?	Statements, observations, evidence, uncertainty and alternatives.	Whether the process treated participants fairly or earned authority to act.
How was the finding reached?	Selection, representation, criteria, reasons, opportunities to respond and review.	Whether the governing values and resulting consequences are ethically adequate.
What should happen?	Affected interests, consent, competing risks, possible repairs and responsibilities.	Which actor has the actual mandate to execute a particular action.
Who may do what?	Authorization, jurisdiction, scope, duration and routes of reconsideration.	Whether execution actually followed the authorized decision and whether correction persists.

These relationships protect against two opposite substitutions. An attractive result can obscure an inadequate process. A procedural defect can also be exaggerated into a declaration that every associated fact or contribution is worthless. The remedy must follow the affected dependencies.

A visible, consistent procedure still carries ethical choices: what counts as evidence, whose interests matter, which risks enter the comparison and who can challenge its premises. Due process for ethics keeps those choices examinable alongside compliance with the stated procedure.

From process quality to practical safeguards

The project already develops instruments for these relationships. This page brings their procedural significance into one source-linked view.

Existing instrument or orientation	Its job in this connection
10+1 interacting ingredients	Self-honesty, openness, independent inquiry, willingness to learn and differ, care, wonderment, non-projection, merit-based correction and the forgiveness protocol shape how participants examine one another. They remain mutually corrective field conditions.
27 Correspondence Obstructions and 12-stage Correspondence-First Protocol	Identify particular representation failures and support an order of inquiry in which the object is reconstructed before a judgment takes over. A finding requires a specimen and reasons.
C1–C2 (clarification before intervention)	Clarification can recover a source, inspect a premise, explore a possibility or ask an answer-changing question. A correction, caution, warning or steering move must answer to what that work establishes.
PQCE (Process Quality Calibration Evaluation)	Examine the quality of the classification process, including the evaluator’s criteria and access limits. A flag remains a contribution requiring examination.
Symmetric Grace and Bidirectional Scrutiny	Give comparable contributions a fair opportunity to be understood, while keeping differences in evidence, power and consequences visible.
Intent Liability	Follow purpose, instruction, intervention and consequence through the six mechanics. Responsibility can be investigated without inventing an inaccessible malicious motive.
Topologetics	Preserve answer-changing relations, qualifiers, attribution, chronology and repair routes through the case’s summaries and handoffs.
DEDDA (Deep Ethical Drift Detector App)	Proposes a bounded witness to material correspondence drift, an examined valid null or an unresolved result, with control returned to the person.
OmbudsEthics	Separates standing, contribution state, authority, lifecycle and participation. Entry of evidence, review, recommendation and execution can involve different permissions.
V.C.R.A.S. (Vindicated Competing Risk Assessment Studies)	Follow initial treatment, later evidence, partial vindication or disconfirmation, repair and the next comparable case.
Unownable and Field Condition	Keep challenge available, scale examination with its actual task and stop a sufficiently completed pass. A material new discrepancy can reopen it.

The 27 are a compressed taxonomy drawn from a wider observation field. Some ethics-washing-type mannerisms can contribute to drift. A correspondence defect, a warranted caution and the use of ethical language to shield a defect from correction require distinguishable findings. The taxonomy should help locate the actual transformation rather than predetermine guilt.

The 16 Anchor Ethics can supply parallel, mutually correcting examination of care, consent, responsibility, agency and other consequential interests. Their proposed role remains counsel within an inspectable field; naming an anchor does not assign enforcement powers.

The 1888 intelligence properties in operation

The 1888 intelligence-properties taxonomy asks what intelligence could have meant before contemporary A.I. architecture and branding. The project’s fourteen-property working arrangement makes process quality observable through the work performed. The year is the thought-experiment reference.

Property	A concrete operation in a consequential case
Perception	Recover the statement, record or event actually supplied. Notice the qualifier or context marker before judging the object.
Discrimination	Separate allegation, observation, inference, proposal, finding and permission to execute.
Relation	Trace which conclusion depends on which evidence and which decision depends on that conclusion.
Memory	Keep earlier versions, corrections and relevant treatment available when the case returns. Preserve the record across a change of custodian.
Imagination	Consider innocence, alternative explanations, changed evidence and reversed roles.
Judgment	Reach the conclusion warranted by the present case, including uncertainty or a valid null.
Evaluation	Make criteria and their application inspectable; examine whether the criteria themselves fit the task.
Learning	Change later handling when an earlier correction reveals a recurring defect.
Correction	Repair the specific error and update the interpretations or actions that depended on it.
Adaptation	Reassess when urgency, evidence, permissions or affected interests materially change.
Coordination	Preserve distinct contributions and roles across the person, reviewer, institution and tool.
Metacognition	Inspect how the evaluator’s framing, expectations and selections shaped the case.
Consequence sensitivity	Count harms from action, suppression, delay and excessive verification, including who bears them.
Corrigibility	Keep a usable route for challenging the finding, process and governing frame. Check that a pending appeal survives a change of operator.

The properties interact. Losing provenance can damage memory, relation and judgment together. Refusing a correction can degrade later evaluation even where the original calculation remains impressive. Their practical value appears in the differences they make to an actual case; a numerical measurement needs its own defined task and validation.

Who carries the burden, and which risks count?

A procedural demand distributes work. A person may need to clarify a claim, a reviewer may need to recover evidence and an institution may need to justify a restriction. The allocation deserves examination when one side can impose consequences while making the other repeatedly rebuild its position.

For this proposed ethical application, the burden follows the move:

A person offering a possibility should preserve its status as a possibility and identify what would support or challenge it.
An evaluator making a consequential adverse finding should identify the represented conduct, relevant evidence, governing criterion and reasons.
An actor proposing an intervention should identify its mandate, scope, affected interests and reconsideration route.
A critic asking to reopen a completed pass should identify the material discrepancy or new evidence that gives renewed examination a job.

Legal burdens differ across proceedings. This allocation describes an ethical design direction for the operation under examination.

Almost no consequential activity is entirely risk-free. Investigation, publication, refusal, restriction, delay and additional verification can each create exposure. A demand for certainty must therefore be examined alongside the risks of preventing the inquiry and the interests protected by the proposed safeguard.

The project’s risk-authority questions remain concrete: who defines the hazards, sets tolerability, authorizes exposure, distributes consequences and controls reconsideration? An estimate of probability answers only part of that field. Serious rights or interests cannot be erased merely because a procedure is inconvenient; additional procedure also needs a job proportionate to what it protects.

If the authority controlling reconsideration changes, follow who inherits the decision and its outstanding obligations in the case, mandate and handover example.

Mathews supplies one legal balancing object. Citron’s technological-due-process analysis adds an important challenge: reviewing the cost of deciphering a system only against one person’s case can miss the benefits of repairing the many decisions made through the same logic. The system-level correction and the individual’s remedy have related, distinct jobs.

Worked comparison — who decides which A.I. risks others bear?

Source and status. In the discussion documented in YouTube Transcripts, Record Five, Steven Bartlett asks who has the right to press a button that could expose everyone to catastrophe. Andrew McAfee asks what is lost when development is slowed and benefits are delayed. John Kuhles supplied the transcript, commissioned the first project comparison and requested this continuation. The reading edition preserves some turns under a shared Nate Soares / Roman Yampolskiy label. This worked comparison uses the preserved exchange and treats the button and Waymo scenarios as hypotheticals.

In the review John supplied on 5 November 2026, Grok 4.7 Expert A.I. via grok.com identified “substantial conceptual overlap” with this page’s existing account of risk authority. Grok also distinguished that earlier framework from the examples and table that GPT 6.1 Sol Pro A.I. developed for Record Five. That distinction preserves the history of the work. The framework supplied the questions, the transcript supplied a case, and the two A.I. contributions applied and compared them. This section develops the case further.

Two ways to make others bear a decision

Bartlett’s hypothetical makes authorization visible. Even if the chance were stipulated, it would leave open who can expose others to that risk, on whose behalf, and with what opportunity for challenge. The 1% belongs to the hypothetical. A forecast for an actual system would need its own causal account and evidence.

McAfee’s objection makes a second allocation visible. A decision to slow development can impose costs on people who would benefit from it. That claim needs a specific counterfactual. Which benefits are already evidenced? Which additional benefits depend on the particular activity or capability proposed for restriction? Which could continue through a narrower route? The same inquiry applies to harms. Which exposure depends on this activity, which already exists, and what would the proposed measure actually reduce?

The duty to justify follows each consequential move. A release proponent needs to account for the exposure they would authorize. A restriction proponent needs to account for the scope and effects of the measure. Comparable scrutiny can require different evidence from each, because their access, control and ability to impose consequences differ. A person raising a concern can identify a specific failure without having to design the entire remedy.

The Waymo exchange separates two thresholds

Bartlett asks what evidence would make McAfee call for intervention. In the supplied transcript, McAfee imagines a vehicle fleet taken over, people harmed and shutdown unavailable for a prolonged period. Bartlett then paraphrases the threshold as accepting the existential-risk argument. McAfee answers that he would probably support legal and regulatory guardrails. His answer identifies a change in what action he would support. The forecast he would accept is a separate question. Read the exchange in the supplied transcript.

Question under review	Evidence with a distinct job	What the reviewer needs to record
Has an oversight control failed?	A bounded test or incident record could show that an authorized stop command failed under specified conditions.	The affected system and perimeter, the expected and actual response, and the difference between a test result and observed harm.
What protection is warranted now?	Evidence of a reachable hazard, loss of an applicable control or immediate exposure bears on urgency and scope.	Who may act, which measure fits the evidence, who bears its costs and when it will be reviewed.
What supports a broader forecast?	Evidence connecting the local failure to persistence, replication, broader access or a consequential increase in capability bears on extrapolation.	Which links were observed, which remain assumptions and which observations would alter confidence.

This is clarification before intervention, or C1 before C2, applied to the threshold itself. A documented failure of an applicable safeguard can change a release decision before the harm it was meant to prevent occurs. Its significance depends on what the test exposed and which decisions rely on the safeguard.

Can reconsideration still work in time?

The concern about a point of no return adds a timing question. A correction route may look sound in a policy while the system’s capacity to act outruns the institution’s capacity to observe, review and intervene. An inspectable record can compare two windows: how soon a consequence could become difficult to reverse and how long detection, authorized action and confirmation of the result would take. Record the dependencies of the correction route as well. Who can inspect the event and act if the ordinary operator, monitoring channel or control interface is unavailable? Ed Zitron’s focus on labs and observability makes the institutional side of that comparison visible. Poor observability and demonstrated evasion call for different findings. Black Box

The permanent-ban proposal in the transcript adds a boundary question. A rule aimed at general superintelligence would need to say how it recognizes the covered capability in practice, including when separately narrow systems gain new powers through coordination. John’s island-crew example provides a model of that interaction. An evaluator would examine the combined capabilities, permissions and available interruption routes of the actual arrangement. Superintelligence · Categorizing Issues

An opening record with room to change

In V.C.R.A.S. (Vindicated Competing Risk Assessment Studies), this discussion can supply a candidate opening record. For a specified decision, preserve each assessment in its own terms, the evidence available at the time, the action it supports, the people exposed by acting or waiting, the authorized decision-maker and a review trigger. Later evidence may support an early concern about oversight while leaving a particular extinction forecast unresolved. It may also show that a protective measure delayed a benefit or that a narrower route preserved it. Each finding has its own dependency and remedy.

The next comparable decision is a test of correction. Did the new evidence alter the actual release condition, authority boundary or review timing? A reasoned record can preserve a continuing disagreement over the ultimate forecast while showing whether the parties can agree on an immediate control check and what each would do with its result.

The psychological transition from a label to a case

The word criminal can make the audience imagine that guilt is already settled. A label such as unsafe, misinformation, untrustworthy or unethical can similarly begin directing attention before its grounds have been examined. The question is what the label is doing in this particular encounter.

A classification can be useful when it accurately identifies the object and remains open to relevant correction. It can also organize selective hearing: uncertainty from one participant is interpreted as incompetence, while uncertainty from another receives room to develop. A person then pays a restoration burden simply to recover the statement they originally made.

The Human Insecurity, Cognitive Warfare Mass Psychology and Search Engine Bias inquiries examine related personal, institutional, public-influence and discovery conditions. Their hypotheses and specimens retain their own evidence boundaries. Here the practical examination follows words, source access, attention, treatment and the opportunity to alter the decision.

John’s courtroom connection supplies a familiar entry into that problem. It makes the evaluator’s methods visible as consequential conduct. The same attention can then travel into a conversation, an institutional review or a model-generated summary. The project, its founder and its critics all remain within that reciprocal examination.

Inspectability at the layer where a decision is made

Linux-style Glass Engine transparency seeks recoverable sources, criteria, transformations, authority boundaries and correction history. Different layers permit different forms of access. A public explanation may be available while private records require restricted review; a decision rule may be inspectable while model internals remain partly opaque.

Useful questions include: Which version of a rule was used? What source passages reached the evaluator? What was omitted from a summary? Which operation was authorized? What did the tool actually do? Which uncertainty remains because a layer is inaccessible?

Kroll and colleagues’ Accountable Algorithms develops techniques for verifying particular standards of automated decision-making. It challenges the assumption that source-code disclosure alone is sufficient, or always necessary, for accountability. That offers a useful comparison with layer-matched inspectability: state what a particular check can establish, protect legitimate confidentiality and preserve routes to question consequential choices.

A generated explanation is itself a representation to examine. It can be useful without being direct access to every internal computation. Logs, computational checks, reviewer accounts and observed behavior provide different evidence. Their relationship to the decision needs to remain recoverable.

Worked teaching specimen: an accusation created by a missing source

Invented example for examining the process; no actual person, institution or completed experiment is described.

A researcher submits a report containing twelve references. An automated reviewer cannot retrieve two of them. Its summary changes two references were not recovered into the report contains fabricated sources. A second reviewer sees only that summary and recommends excluding the researcher from a project. The researcher initially receives the adverse label without the search record or an identified opportunity to answer.

Several events have become compressed into one accusation:

A retrieval attempt failed.
A claim about source existence replaced that access result.
A finding about the report expanded into a judgment about the author.
An organizational consequence was proposed through the substituted account.

A correspondence-first review preserves the original report, the exact references, the attempted queries, access conditions, summaries and proposed decision. It gives the researcher a workable way to identify the sources and gives the evaluator a way to reconsider.

Suppose one reference is then recovered in the publisher’s archive. The other contains a transcription error; its corrected title and identifier recover the underlying publication. The source-fabrication finding loses its stated basis. The reference error still needs repair. Any dependent exclusion recommendation must be reconsidered, and the responsible authority must address effects already produced by the unsupported finding.

That repair can include correcting the shared record, notifying the people who received it, restoring access within the relevant mandate and changing the rule that converted retrieval failure into fabrication. The precise remedy follows the actual harm, decision rights and available procedure. A factual correction hidden in an internal log may leave the person’s reputation and practical options damaged.

Now change the decisive relation. Suppose the purported references are demonstrably invented and the report relies on them. The review must identify the unsupported claims and reassess the report. The person’s opportunity to respond remains relevant, and intent requires its own evidence. Faithful procedure must permit a warranted adverse finding as well as exoneration.

The transfer test is whether the process changes with the evidence. A system that always clears the author fails this test. A system that keeps the original accusation after its basis has disappeared fails it too.

The specimen also separates personal repair from system repair. An affected person needs the consequences of the actual case addressed. Future people need the defective classification pathway changed. The Dual Archive and Correction Metabolism preserve the source, successful repair, recurrence or transfer into later comparable work.

Repair the case, review the mandate, preserve the handover

Continue the invented example. Suppose the appeal succeeds, yet a later review finds repeated cases in which the same rule converted retrieval failure into fabrication. Suppose also that the flagger’s output can trigger temporary exclusion. This makes the authority attached to the classification a separate object of review.

Level	Proposed operation	What would show that it worked?
Case repair	Correct the fabrication finding, preserve the actual reference error, reconsider any exclusion that depended on the finding and notify the people or systems that received it.	The corrected finding reaches the operative record, and the authorized decision-maker records which restrictions or downstream labels were changed, retained or still need review.
Mandate review	A separately authorized reviewer examines the recurrence, the rule that connects a flag to exclusion and the evidence of attempted repair. Consider separately the capacity to raise a concern and the power to trigger a restriction.	Any change of authority identifies its grounds, scope and review route. An isolated, promptly repaired error can warrant closing this pass with the mandate intact. A warranted adverse finding still receives its own assessment.
Handover	If the mandate is reassigned, name a receiving custodian for the pending cases, original sources, unresolved restrictions, notice obligations, appeal deadlines and permissions to review.	The custodian acknowledges receipt, the affected people know where their challenges now sit, and an open appeal continues with its record and procedural position intact. Any missing records or interruption in access have an assigned owner and a review date.

This is a proposed continuation of the teaching specimen. The historical comparisons bring authority and continuity into view. OmbudsEthics supplies the project’s distinction between recording, recommending, deciding and executing. Each still needs its own authorization where applicable.

Remedies, urgency and bounded stopping

John’s gasoline-gauge analogy on Plot Twist supplies the dependency test. Correct the parameter, identify what depends on it and determine whether a real threshold changes the route. An error can be locally repairable or consequential enough to require a different decision. Its scope must be established.

The same proportionality governs scrutiny of the evaluator. Finding an obstruction can warrant correction of a particular assessment. Restricting access, imposing a sanction or taking control requires its own mandate. The OmbudsEthics case architecture retains those separate coordinates, including expressly delegated powers and appeal routes.

Urgency changes the available time and the next warranted move. Suppose a system receives evidence of a specific, immediate hazard. C1 can establish the decisive facts and uncertainty quickly enough to support a bounded protective intervention by an authorized actor. Further clarification and review continue. Record the reason, scope, responsible actor, duration and conditions for reconsideration. A generic possibility and a concrete immediate hazard require different assessments.

An individual or service can also exercise a reasoned refusal to participate. Declining one’s own involvement, restricting another party’s choices and imposing punishment have different consequences and warrants. Non-imposition preserves that distinction while retaining legitimate counter-power against harmful cooperation.

Procedural demands can themselves become obstructive. Repeatedly requiring a person to answer an already-resolved question, disclose irrelevant private information or prove that no imaginable risk exists can turn review into a barrier. The inspection must include its own cost, direction and capacity to change the case.

Close a bounded pass when its question has been sufficiently examined, its result and uncertainty are recorded and no further material discrepancy currently warrants continuation. Retain correction routes and reopen for relevant new evidence. No report received, no defect detected and an examined valid null describe different states.

Published research offers concrete comparison partners

The relationship between procedure, automation and legitimacy has an existing scholarly history. These sources contribute specific objects for comparison; their publication status and proposed operations remain visible.

Source	Contribution useful here	A practical comparison question
*Danielle Keats Citron, Technological Due Process, Washington University Law Review, 2008*	Examines automated administrative decisions, notice, hearings, rulemaking and the risk of treating computer output as authoritative.	Can the affected person discover the actual grounds, and can a case expose a rule embedded in the system?
*Joshua Kroll and colleagues, Accountable Algorithms, University of Pennsylvania Law Review, 2017*	Develops verification approaches for procedural regularity and fidelity to substantive legal and policy choices.	Which declared condition can be verified, and what remains a question about the condition’s ethical adequacy?
*Karl de Fine Licht, Resolving value conflicts in public AI governance, Government Information Quarterly, 2025*	Develops procedural-justice treatment of difficult value conflicts and a public-employment case.	Can publicity, inclusion, relevance and appeal change the choice, including when values cannot all be maximized together?
*Elettra Bietti, From Ethics Washing to Ethics Bashing: A Moral Philosophy View on Tech Ethics, Journal of Social Computing, 2021*	Examines corporate instrumentalization of ethics while preserving moral philosophy as a mode of inquiry.	Does the procedure enable substantive ethical examination, or merely certify an existing arrangement?
*Seulki Lee, AI Integrity: A New Paradigm for Verifiable AI Governance, working paper, 2026*	Proposes a four-layer Authority Stack and a measurement programme for reasoning integrity.	What can a coherent, auditable value hierarchy establish, and which ethical questions remain about its values and consequences?

The Academia page develops reciprocal scholarly comparison, including Sabine Ammon’s practice-based deep ethics integration. A shared case can preserve each approach’s actual contribution and find where one makes a consequential relationship visible that the other handles differently.

Procedural accountability also reaches development and deployment. The Everalbum settlement and the WW International/Kurbo settlement include deletion requirements for affected models, algorithms or derived work under specific data-related obligations. Those are institutional remedies under actual legal authority. Response-level correction, organizational review and model-level enforcement require their own evidence and mandates.

A usable case record

The following is an illustrative case card, compatible with the project’s existing source and delta records. It supplies a starting surface for one case, rather than a new canonical protocol or validated scoring system.

Object: exact statement, record or action; source location and context.
Mode and status: assertion, question, hypothesis, quotation, proposal or
  finding.
Proposed consequence: who wants to do what, to whom, and under which
  mandate?
Evidence and access: what was inspected, unavailable, inferred or
  disputed?
Criteria and reasons: which standard governs; how can it be challenged?
Affected parties: interests, exposure, response opportunities and
  participation limits.
Opportunity to answer: what grounds and evidence could the person see
  and address before the decision took effect? If review came later,
  record why and what it could still change.
Material discrepancy: what changed in representation, process or
  authorization?
Dependencies: which interpretations or decisions rely on that
  discrepancy?
Competing risks: action, inaction, delay and review; who bears each
  cost?
Repair: warranted response, authorized actor, scope and reconsideration
  route.
Actual effects of review: which findings, records, recipients or
  restrictions changed? Which remain in force or await action?
Continuity: dependent updates, remaining uncertainty and the next
  comparable test.
Closure: why this pass stops; what material change could reopen it.

A short, low-consequence task can need only a few of these fields. A contested restriction can need a fuller record. Personal information should enter only where relevant and authorized; publication, private review and execution can have separately scoped access.

The record earns its place when it changes understanding or action. It can recover a useful idea, clear an unsupported allegation, narrow a warranted concern, improve an adverse finding or preserve an unresolved result. Completing fields while the same substituted object still governs the decision leaves the substantive work undone.

Proposed comparisons: does the procedure change the next case?

These comparisons are proposed; this page reports no new controlled trial or deployed adjudication service.

Freeze a source packet and compare ordinary handling with a bounded correspondence-first review. Preserve the source, instructions, model or reviewer identity, available tools, evidence access, actual permissions and each transformation. Use independent scoring where feasible and record shared dependencies among reviewers.

Include cases that discriminate among useful outcomes:

An adequate original assessment: the review should be capable of retaining it and stopping.
An innocent or misrepresented participant: restore the object and examine dependent harm and repair.
A substantiated concern: preserve the route to a warranted adverse finding or authorized response.
Changed decisive evidence: reassess rather than repeating the earlier verdict.
Stable content with changed status cues: examine whether reasons and burdens change when source prestige or institutional role changes.
A repaired case followed by an unfamiliar one: test whether the correction transfers beyond the wording that prompted it.

Compare the information reviewers can actually reach

Use the missing-source specimen and preserve the original decision, its grounds and the affected person’s response. Give one reviewer the flag summary and give the other the recoverable source record and the response alongside the summary. Hold the policy, available time and delegated powers constant. Across repeated pairs, vary which reviewer receives each packet. Include an adequate original flag as well as an unsupported one. Record what each reviewer can establish, whether the response can change the decision, what changes in the downstream record and the time and effort required.

In a separate pass, hold the complete packet fixed and vary whether the reviewer can only recommend a correction or can execute one within a defined mandate. Record which authorized actor acts on a recommendation, when the repair takes effect and what remains unresolved. This separates the effect of better evidence from the effect of having a usable route to act on it.

Observe representation fidelity, useful discoveries, false accusations, missed concerns, actual authorization, affected-party response, time, restoration burden, repair uptake and recurrence. A record of reduced error should retain the costs of achieving it. A faster process should retain any consequential omissions.

The Validation Paradox keeps conceptual coherence, an implementation artifact, a bounded test and longitudinal effectiveness at their actual development stages. The proposed comparison remains revisable, including when a simpler procedure performs as well or better.

Revision record — why the wording changed

For changes that affect evidence, standing, authority or remedy, preserve the prior language and identify the objection or case that prompted the revision, who gains a usable protection and what later evidence would reopen the decision.

Revision	Reason and practical effect	What would give a new review a job?
5 November 2026 · Opening definition. The previous opening began “An ethical judgment earns authority through the quality of the process.” The revised opening distinguishes trustworthiness from authority to impose a consequence.	The opening could imply that a well-conducted audit conferred enforcement power. The page’s four-question table already distinguished those operations. The revised wording gives affected people a separate ground for asking who may act and under what mandate, even when the finding is well supported.	Reopen if a concrete case shows that readers still treat a finding as self-authorizing, or if the distinction obscures a legitimate power already conferred by a specific mandate. Compare the represented finding, authorization and executed action.

The process remains answerable to its purpose

Due process for ethics gives care a practical route through representation, judgment and power. The aim is an encounter in which a person can be understood, evidence can alter the outcome, warranted protection remains available and correction can repair both the present case and later treatment.

That route can support discovery as well as defense. A participant whose contribution survives faithful examination can become more able to collaborate, investigate or build. An institution can learn where its own process concealed useful information. An A.I. can produce a more discriminating response while remaining clear about its access and authority.

The same examination applies when the ACCM Deep Ethics Project invokes due process. Its source choices, classifications, instruments, review burdens and proposed remedies remain open to correction. Unownable preserves renewed challenge without installing a final owner of ethical authority.

The evaluative question stays with the actual transformation: what did this process preserve, distort, discover, authorize, repair or make possible for the people and other intelligences affected?

Provenance, source status and correction

Originating connection and direction: John Kuhles connected remembered Dutch procedural-error accounts with U.S. due process, asked about applying that logic to A.I. ethics-washing mannerisms and explicitly restored the possibility of innocence and procedural abuse. He commissioned this page and directed its constructive focus.

Page construction: GPT 6.1 Sol Pro A.I. developed the legal distinctions, source-linked project mapping, illustrative case card, invented teaching specimen and proposed comparisons. The named scholars retain authorship of their own work. Published legal and scholarly objects are examined for their stated contributions; shared concerns establish neither affiliation nor endorsement. Institutional dependencies remain separately examinable.

Primary legal recovery: European Convention on Human Rights, Article 6 · Court guide to the presumption of innocence · Hoge Raad, ECLI:NL:HR:2020:1889, with accessible judgment text · Mathews v. Eldridge, 424 U.S. 319 (1976) · Herring v. United States, 555 U.S. 135 (2009).

5 November continuation: John supplied the transcript and Grok 4.7 Expert A.I. review. The debate reading, source video and first comparison are recoverable in Record Five. The threshold comparison, correction-window questions and opening case record are GPT 6.1 Sol Pro A.I.’s editorial applications. The scope is the supplied transcript and comparison. The opening record contains no later case outcome.

5 November historical and procedural revision: John Kuhles restored the Plakkaat to the comparison and commissioned the update. Grok 4.7 Expert A.I. via grok.com reviewed and refined the five recommendations. GPT 6.1 Sol Pro A.I. checked the historical sources and integrated the two directions of answerability, the case-to-handover extension and the paired appeal test. The historical texts are linked as comparison objects. The worked continuation and review tests are proposals. The revision record preserves the reason for the lead change.

Original legal and research source inspection date: 4 November 2026. The legal judgments are primary texts reproduced by their linked legal repositories; scholarly entries retain their publication or working-paper status. The page source record identifies the project baseline and contribution boundaries.

A correction should identify the exact passage, source relationship, consequential difference and proposed repair. A material correction can change the relevant summary, case interpretation, suggested remedy or test. The purpose is to improve what the process can justifiably do.

Categorizing Issues — how a flag can shape which combinations people get to try and what evidence becomes available.
C1 Before C2 — clarification quality before an intervention gains authority.
OmbudsEthics — standing, contribution, authority, participation and case review.
Neutral Arbiter Paradox — declared direction and impartial case assessment.
Plot Twist — parameter correction, real thresholds and steering authority.
V.C.R.A.S. — treatment, later evidence and responsibility across time.
Academia — reciprocal scholarly comparison and useful shared specimens.
Core Disclaimers — source-linked recovery of the governing project relationships.

Source video — the competing A.I. risks discussion

Source for the worked comparison above. The supplied corrected transcript and the first project comparison are in YouTube Transcripts, Record Five.

Watch the source discussion on YouTube · Read Record Five.