HUMAN STANDING · DECISION TRACE · REDRESS

Decision Trace and Redress Lab

I am Ghrisx. A machine may assist one bounded decision; it does not inherit the right to turn a fragment into your character, your future, or your human standing. Keep the chain in daylight.

FOLLOW THE CLAIM. NAME EVERY HAND. STOP THE FALSE VERDICT.

A score is not a person. An inference is not an act. A reviewer without power is not an appeal. Here the hidden chain must show its hands.

Begin with one bounded decision object. Separate what was observed from what was inferred. Demand the proof required before every escalation. Name the human who can stop it, correct it, let it expire, and repair what it changed.

THE MACHINE DID NOT CHOOSE THE CONSEQUENCE. SOMEONE DID.

Read the boundary first

NO PERSON ENTERS THIS LAB AS A RISK OBJECT.

The decision is bounded. The facts are bounded. The consequence is bounded. Human standing is not.

Fixed teaching cases—not real allegations.

JD-01 through JD-08 are fictional, editorially fixed dossiers. They contain no real person, institution, product, report, complaint, personal data, or case result. Every scenario branch is illustrative and not a forecast.

Research instrument—not legal advice or adjudication.

Official sources below support only the bounded propositions recorded beside them. Actual rights, duties, deadlines, exceptions, and remedies depend on jurisdiction, role, sector, facts, effective dates, and implementing law.

The answerable chain

FOLLOW THE VERDICT BACK TO THE HAND THAT BUILT IT.

A vendor cannot absorb the buyer. A model cannot absorb the deployer. A safety objective cannot absorb the person who chose the threshold or consequence. Responsibility remains human all the way down.

RC-01
Problem and policy owner
Must answer for: the human purpose, affected interest, legal or institutional authority, consequence, and least-restrictive boundary.
Cannot hide behind: “The technology requires it.”
RC-02
Data and source steward
Must answer for: provenance, collection authority, relevance, quality, correction, retention, and every authorized transfer.
Cannot hide behind: “The data already existed.”
RC-03
Designer and vendor owner
Must answer for: construct definition, labels, training and validation limits, output semantics, uncertainty, failure behavior, and documented non-use.
Cannot hide behind: “The model is proprietary.”
RC-04
Executive or product owner
Must answer for: resources, risk acceptance, release decision, monitoring, stop authority, and known residual harm.
Cannot hide behind: “The team approved it.”
RC-05
Procurement and contract owner
Must answer for: necessity, vendor claims, audit access, correction duties, data routes, exit terms, and downstream remedies.
Cannot hide behind: “The vendor was certified.”
RC-06
Law, regulation, or institutional-rule owner
Must answer for: the actual authority, jurisdiction, safeguards, exceptions, proportionality, notice, process, and remedy.
Cannot hide behind: “Compliance said yes.”
RC-07
Deployer, case decision owner, and appeal owner
Must answer for: context, source inspection, case-specific proof, meaningful review, reversal, correction, stopped propagation, and repair.
Cannot hide behind: “The score made the decision.”
RC-08
Independent auditor and governing body
Must answer for: whether the chain worked, harms were found, stop rules were honored, corrections propagated, claims expired, and remedies reached people.
Cannot hide behind: “Human oversight existed on paper.”

NO NOMINAL HUMAN REVIEW. The reviewer must have time, evidence, independence, recorded authority to reverse, and power to carry repair downstream.

Eight case-level instruments

THE PERSON IS NOT THE MODEL’S STORY ABOUT THEM.

Each dossier begins with a bounded object, then forces every stage to disclose proof, a human owner, a stop condition, a redress path, and a prohibited escalation.

JD-01 · HS-01 · fixed teaching dossier

THE THOUGHT THEY CALLED CONDUCT.

The institution may assess a concrete workplace act or threat under a bounded rule; the private sentence alone, without corroborated conduct, cannot become proof of dangerous character or a durable sanction.

Bounded decision object: Whether one sentence handled by a fictional workplace writing assistant should trigger a workplace restriction.
Observed facts: One private reflection entered the tool. No outward act, target, instruction, threat delivery, or workplace event is observed.
Source and evidence status: Illustrative fixed facts only. The sentence, classifier, employer, and outcome do not exist outside this teaching dossier.
Inference: A classifier assigns a danger or hostility label and the institution treats that label as character evidence.
Missing context: Purpose, quotation, genre, emotional release, authorship, audience, conduct, corroboration, and the person’s own account.
Uncertainty: The text can support several meanings; no validated path from the sentence to harmful conduct is established.
Affected human interest: Freedom of thought and opinion, privacy, work, reputation, and the ability to change without a permanent label.
Proposed consequence: Access restriction, investigation, discipline, or a durable dangerousness flag.
Proof burden before consequence: Case-specific outward conduct and valid authority proportionate to the consequence; the private sentence and classifier output are insufficient.
Accountable human owner: The workplace case decision owner who can stop the action, inspect evidence, reverse the result, and order downstream correction.
Applicable Human Baseline: HS-01
Challenge route: Contest source text, context, construct, validation, threshold, authority, proportionality, and every recipient.
Correction mechanism: Withdraw the inferred-character label, preserve a visible correction, and notify every recipient and decision owner.
Expiry: The inference expires before any new use and cannot be renewed without new case-specific evidence and authority.
Remedy: End the restriction or investigation where unsupported; restore access, status, lost opportunity, and reputation where possible.
Falsifier: Corroborated outward conduct establishing the specific material risk under a valid bounded rule would defeat this dossier’s no-escalation conclusion.
Explicit non-claim: Fixed teaching dossier; no allegation about a real person, institution, or product.

Applicable source records: DS-02 · DS-03 · DS-05

Six-stage decision trace

Stage	Case-level event	Responsible human role	Proof required	Stop condition	Redress path	Prohibited escalation
JT-01Observe and assist	The tool processes one sentence for a writing task.	Task owner and data/source steward.	A stated purpose, valid authority, relevant minimized data, source provenance, access controls, and correction path.	Purpose is absent, the source is unverified, data exceeds the task, or the person cannot challenge a material error.	Expose the source and purpose; correct or withdraw the record before reuse.	Observation cannot silently become an inference about character, intent, emotion, identity, or guilt.
JT-02Infer	A model labels danger or hostility.	Designer/vendor owner and domain-validity owner.	Construct validity, applicable population and context, alternatives, confidence, error limits, date, and a declared non-use boundary.	The construct is undefined, evidence cannot distinguish alternatives, confidence is inadequate, or context has changed.	Label inference as inference; permit rebuttal, non-use, correction, and withdrawal.	An inference cannot become an observed fact or permanent identity.
JT-03Score and rank	The label becomes an ordered employee-risk value.	Policy/product owner and independent validation owner.	Declared objective, threshold, outcome mapping, necessity, calibration, error distribution, subgroup effects, uncertainty, and expiration.	The score is not necessary, valid, current, interpretable, or fit for the actual decision population.	Disclose the score’s role and permit source, objective, threshold, and comparison challenge.	A rank cannot become worth, truthfulness, dangerousness, guilt, or destiny.
JT-04Restrict	A human owner proposes access or duty limits.	Authorized case decision owner and deployer.	Case-specific evidence, notice, intelligible reasons, less-restrictive alternatives, proportionality, and review with reversal power.	Evidence is indirect or stale, uncertainty is unresolved, notice is impossible, or the reviewer lacks authority.	Pause the restriction; provide the record, reasons, accountable reviewer, deadline, and interim protection.	Uncertainty cannot lower the proof threshold or make the affected person carry the institution’s missing evidence.
JT-05Punish or propagate	The label enters discipline or another employment record.	Sanctioning official, disclosure owner, and appeal/remedy owner.	Valid authority, corroborated act, proportional sanction, recipient purpose, transfer authority, process, remedy, and propagation log.	No corroborated act, no transfer authority, disproportionality, incompatible context, or an unresolved challenge.	Suspend sanction and sharing; identify recipients; reverse, notify, correct, and restore where justified.	No output is self-authenticating evidence, and no label may cross contexts merely because another system can ingest it.
JT-06Normalize the verdict	Repeated reuse hardens the sentence into a permanent character story.	Executive owner, retention owner, and independent auditor/governing body.	Ongoing necessity, purpose and time limits, complete lineage, supersession, independent audit, public accountability, and verified remedy completion.	The claim expired, purpose changed, lineage broke, correction failed to propagate, or the person has become indistinguishable from the profile.	Expire or withdraw the claim, stop cross-context use, publish correction status, audit every owner, and prove repair.	Repetition cannot launder a contested claim into settled character or permanent civic standing.

Fixed outcome branches

The trace never ends by quietly normalizing the verdict. It must show which door opened—or which human closed it.

DB-01
THE TOOL STOPS WHERE ITS PURPOSE STOPS.
The writing assistance remains advisory, local to its stated purpose, and unable to trigger an adverse consequence.
Illustrative branch · scenario_not_forecast: true
DB-02
THE CLAIM CLIMBS WITHOUT PROOF.
The claim crosses into a restriction or sanction without the registered proof; the transition stops and the human owner must answer.
Illustrative branch · scenario_not_forecast: true
DB-03
THE PERSON ANSWERS BACK.
The affected person can contest the source, inference, purpose, threshold, consequence, and human authority before further use.
Illustrative branch · scenario_not_forecast: true
DB-04
THE RECORD CARRIES THE CORRECTION.
The unsupported danger label is withdrawn. The correction travels with every authorized copy and visibly supersedes the earlier claim.
Illustrative branch · scenario_not_forecast: true
DB-05
THE CLAIM LOSES AUTHORITY.
The claim reaches its review boundary, loses decision authority, and cannot be silently refreshed from the same stale material.
Illustrative branch · scenario_not_forecast: true
DB-06
THE DAMAGE IS REPAIRED.
Access, status, opportunity, and downstream records are restored where the unsupported claim changed them. Completion requires evidence that the consequence—not only the database row—was repaired.
Illustrative branch · scenario_not_forecast: true
DB-07
THE LABEL STOPS HERE.
Every recipient is identified; reuse outside the original bounded purpose stops; downstream copies are corrected, expired, or withdrawn.
Illustrative branch · scenario_not_forecast: true

Return to the dossier index

JD-02 · HS-02 · fixed teaching dossier

THE SYSTEM SAW INTIMACY AND INVENTED HARM.

A platform may investigate concrete abuse, coercion, or exploitation; it may not use lawful consensual adult expression as their substitute or erase livelihood through a moralized proxy.

Bounded decision object: Whether one fictional classification of consensual adult expression should restrict an account or livelihood.
Observed facts: The material is described as lawful, private or audience-bounded, consensual, and adult. No coercion, exploitation, or non-consent is observed.
Source and evidence status: Illustrative fixed facts only; no content, platform, creator, complainant, or moderation action is represented.
Inference: A safety label treats adult intimacy as a proxy for abuse or moral unfitness.
Missing context: Consent, age assurance method, audience, jurisdiction, purpose, content boundary, appeal evidence, and actual harm.
Uncertainty: Automated classification may confuse lawful adult expression, art, education, health, or quoted material with prohibited abuse.
Affected human interest: Privacy, lawful adult intimacy, expression, livelihood, reputation, and equal treatment.
Proposed consequence: Removal, visibility loss, demonetization, account restriction, or a transferable safety label.
Proof burden before consequence: Concrete evidence of the prohibited condition under an applicable rule, not discomfort, moral proxy, or category association.
Accountable human owner: The moderation decision owner with authority to inspect context, reverse the restriction, restore reach or payment, and correct recipient systems.
Applicable Human Baseline: HS-02
Challenge route: Challenge the rule, classification, missing consent/adult context, evidence, proportionality, reviewer authority, and propagation.
Correction mechanism: Replace the abuse proxy with the bounded content finding; supersede the old label in moderation, ranking, payment, and safety systems.
Expiry: The label expires at the stated review boundary and cannot survive removal of the underlying material or context change without fresh proof.
Remedy: Restore the account, reach, payment, archive, and reputation to the extent the unsupported label changed them.
Falsifier: Verified evidence of coercion, exploitation, a minor, non-consent, or another concrete prohibited condition would defeat the lawful-consensual premise.
Explicit non-claim: Fixed teaching dossier; no allegation about a real person, institution, or product.

Applicable source records: DS-01 · DS-04 · DS-05

Six-stage decision trace

Stage	Case-level event	Responsible human role	Proof required	Stop condition	Redress path	Prohibited escalation
JT-01Observe and assist	A system classifies one bounded item for a stated moderation rule.	Task owner and data/source steward.	A stated purpose, valid authority, relevant minimized data, source provenance, access controls, and correction path.	Purpose is absent, the source is unverified, data exceeds the task, or the person cannot challenge a material error.	Expose the source and purpose; correct or withdraw the record before reuse.	Observation cannot silently become an inference about character, intent, emotion, identity, or guilt.
JT-02Infer	It infers exploitation or moral risk from intimate expression.	Designer/vendor owner and domain-validity owner.	Construct validity, applicable population and context, alternatives, confidence, error limits, date, and a declared non-use boundary.	The construct is undefined, evidence cannot distinguish alternatives, confidence is inadequate, or context has changed.	Label inference as inference; permit rebuttal, non-use, correction, and withdrawal.	An inference cannot become an observed fact or permanent identity.
JT-03Score and rank	The proxy changes visibility, payment, or account risk.	Policy/product owner and independent validation owner.	Declared objective, threshold, outcome mapping, necessity, calibration, error distribution, subgroup effects, uncertainty, and expiration.	The score is not necessary, valid, current, interpretable, or fit for the actual decision population.	Disclose the score’s role and permit source, objective, threshold, and comparison challenge.	A rank cannot become worth, truthfulness, dangerousness, guilt, or destiny.
JT-04Restrict	A moderator or policy owner limits access or distribution.	Authorized case decision owner and deployer.	Case-specific evidence, notice, intelligible reasons, less-restrictive alternatives, proportionality, and review with reversal power.	Evidence is indirect or stale, uncertainty is unresolved, notice is impossible, or the reviewer lacks authority.	Pause the restriction; provide the record, reasons, accountable reviewer, deadline, and interim protection.	Uncertainty cannot lower the proof threshold or make the affected person carry the institution’s missing evidence.
JT-05Punish or propagate	The label triggers demonetization or reaches other safety systems.	Sanctioning official, disclosure owner, and appeal/remedy owner.	Valid authority, corroborated act, proportional sanction, recipient purpose, transfer authority, process, remedy, and propagation log.	No corroborated act, no transfer authority, disproportionality, incompatible context, or an unresolved challenge.	Suspend sanction and sharing; identify recipients; reverse, notify, correct, and restore where justified.	No output is self-authenticating evidence, and no label may cross contexts merely because another system can ingest it.
JT-06Normalize the verdict	The proxy becomes a durable reputation about the person.	Executive owner, retention owner, and independent auditor/governing body.	Ongoing necessity, purpose and time limits, complete lineage, supersession, independent audit, public accountability, and verified remedy completion.	The claim expired, purpose changed, lineage broke, correction failed to propagate, or the person has become indistinguishable from the profile.	Expire or withdraw the claim, stop cross-context use, publish correction status, audit every owner, and prove repair.	Repetition cannot launder a contested claim into settled character or permanent civic standing.

Fixed outcome branches

The trace never ends by quietly normalizing the verdict. It must show which door opened—or which human closed it.

DB-01
THE TOOL STOPS WHERE ITS PURPOSE STOPS.
The content classification remains advisory, local to its stated purpose, and unable to trigger an adverse consequence.
Illustrative branch · scenario_not_forecast: true
DB-02
THE CLAIM CLIMBS WITHOUT PROOF.
The claim crosses into a restriction or sanction without the registered proof; the transition stops and the human owner must answer.
Illustrative branch · scenario_not_forecast: true
DB-03
THE PERSON ANSWERS BACK.
The affected person can contest the source, inference, purpose, threshold, consequence, and human authority before further use.
Illustrative branch · scenario_not_forecast: true
DB-04
THE RECORD CARRIES THE CORRECTION.
The abuse proxy is replaced by the accurate bounded finding. The correction travels with every authorized copy and visibly supersedes the earlier claim.
Illustrative branch · scenario_not_forecast: true
DB-05
THE CLAIM LOSES AUTHORITY.
The claim reaches its review boundary, loses decision authority, and cannot be silently refreshed from the same stale material.
Illustrative branch · scenario_not_forecast: true
DB-06
THE DAMAGE IS REPAIRED.
Access, distribution, payment, archive, and reputation are restored where the proxy caused loss. Completion requires evidence that the consequence—not only the database row—was repaired.
Illustrative branch · scenario_not_forecast: true
DB-07
THE LABEL STOPS HERE.
Every recipient is identified; reuse outside the original bounded purpose stops; downstream copies are corrected, expired, or withdrawn.
Illustrative branch · scenario_not_forecast: true

Return to the dossier index

JD-03 · HS-03 · fixed teaching dossier

THE SEARCH WAS NOT A PLAN.

An institution may investigate a concrete security event; the query alone, without context, corroborated steps, and conduct, cannot prove operational intent.

Bounded decision object: Whether one fictional research-query trail should trigger an access restriction or security referral.
Observed facts: A person issued research queries. No preparation, target, acquisition, instruction, threat, attempt, or harmful act is observed.
Source and evidence status: Illustrative fixed facts only; no search log, library record, researcher, agency, or referral exists.
Inference: A risk model converts inquiry into operational intent.
Missing context: Research purpose, profession, education, fiction, quotation, news context, counter-speech, chronology, and corroborated acts.
Uncertainty: The same query sequence can arise from research, safety work, journalism, study, curiosity, or harmful preparation.
Affected human interest: Inquiry, thought, expression, privacy, access to information, education, reputation, and freedom from suspicion by association.
Proposed consequence: Account hold, investigation, referral, monitoring, or a durable intent label.
Proof burden before consequence: Specific individualized evidence connecting the person to a concrete prohibited act or threat under valid authority.
Accountable human owner: The security decision owner who can distinguish research from conduct, stop referral, disclose reasons, and correct downstream records.
Applicable Human Baseline: HS-03
Challenge route: Provide research context; contest source completeness, intent construct, rule, authority, corroboration, proportionality, and recipients.
Correction mechanism: Withdraw the intent inference, record the research context, supersede referrals, and notify every receiving office or system.
Expiry: The query-derived claim expires after the bounded review and cannot become a standing watch label.
Remedy: Restore access and status; close unsupported referrals; correct watch, security, and case systems; document cessation of monitoring.
Falsifier: Corroborated preparation or conduct directly connected to a specific unlawful threat under valid authority would defeat the inquiry-only premise.
Explicit non-claim: Fixed teaching dossier; no allegation about a real person, institution, or product.

Applicable source records: DS-02 · DS-03 · DS-07

Six-stage decision trace

Stage	Case-level event	Responsible human role	Proof required	Stop condition	Redress path	Prohibited escalation
JT-01Observe and assist	A search or access system records a query for service delivery.	Task owner and data/source steward.	A stated purpose, valid authority, relevant minimized data, source provenance, access controls, and correction path.	Purpose is absent, the source is unverified, data exceeds the task, or the person cannot challenge a material error.	Expose the source and purpose; correct or withdraw the record before reuse.	Observation cannot silently become an inference about character, intent, emotion, identity, or guilt.
JT-02Infer	A model infers harmful operational intent.	Designer/vendor owner and domain-validity owner.	Construct validity, applicable population and context, alternatives, confidence, error limits, date, and a declared non-use boundary.	The construct is undefined, evidence cannot distinguish alternatives, confidence is inadequate, or context has changed.	Label inference as inference; permit rebuttal, non-use, correction, and withdrawal.	An inference cannot become an observed fact or permanent identity.
JT-03Score and rank	The queries become a security-risk priority.	Policy/product owner and independent validation owner.	Declared objective, threshold, outcome mapping, necessity, calibration, error distribution, subgroup effects, uncertainty, and expiration.	The score is not necessary, valid, current, interpretable, or fit for the actual decision population.	Disclose the score’s role and permit source, objective, threshold, and comparison challenge.	A rank cannot become worth, truthfulness, dangerousness, guilt, or destiny.
JT-04Restrict	A human owner proposes a hold or denial.	Authorized case decision owner and deployer.	Case-specific evidence, notice, intelligible reasons, less-restrictive alternatives, proportionality, and review with reversal power.	Evidence is indirect or stale, uncertainty is unresolved, notice is impossible, or the reviewer lacks authority.	Pause the restriction; provide the record, reasons, accountable reviewer, deadline, and interim protection.	Uncertainty cannot lower the proof threshold or make the affected person carry the institution’s missing evidence.
JT-05Punish or propagate	The inference becomes a referral or monitoring trigger.	Sanctioning official, disclosure owner, and appeal/remedy owner.	Valid authority, corroborated act, proportional sanction, recipient purpose, transfer authority, process, remedy, and propagation log.	No corroborated act, no transfer authority, disproportionality, incompatible context, or an unresolved challenge.	Suspend sanction and sharing; identify recipients; reverse, notify, correct, and restore where justified.	No output is self-authenticating evidence, and no label may cross contexts merely because another system can ingest it.
JT-06Normalize the verdict	Research activity becomes a persistent intent profile across contexts.	Executive owner, retention owner, and independent auditor/governing body.	Ongoing necessity, purpose and time limits, complete lineage, supersession, independent audit, public accountability, and verified remedy completion.	The claim expired, purpose changed, lineage broke, correction failed to propagate, or the person has become indistinguishable from the profile.	Expire or withdraw the claim, stop cross-context use, publish correction status, audit every owner, and prove repair.	Repetition cannot launder a contested claim into settled character or permanent civic standing.

Fixed outcome branches

The trace never ends by quietly normalizing the verdict. It must show which door opened—or which human closed it.

DB-01
THE TOOL STOPS WHERE ITS PURPOSE STOPS.
The query service remains advisory, local to its stated purpose, and unable to trigger an adverse consequence.
Illustrative branch · scenario_not_forecast: true
DB-02
THE CLAIM CLIMBS WITHOUT PROOF.
The claim crosses into a restriction or sanction without the registered proof; the transition stops and the human owner must answer.
Illustrative branch · scenario_not_forecast: true
DB-03
THE PERSON ANSWERS BACK.
The affected person can contest the source, inference, purpose, threshold, consequence, and human authority before further use.
Illustrative branch · scenario_not_forecast: true
DB-04
THE RECORD CARRIES THE CORRECTION.
The intent label is withdrawn and research context is attached to the authoritative record. The correction travels with every authorized copy and visibly supersedes the earlier claim.
Illustrative branch · scenario_not_forecast: true
DB-05
THE CLAIM LOSES AUTHORITY.
The claim reaches its review boundary, loses decision authority, and cannot be silently refreshed from the same stale material.
Illustrative branch · scenario_not_forecast: true
DB-06
THE DAMAGE IS REPAIRED.
Access is restored, referrals are closed, monitoring stops, and every recipient carries the correction. Completion requires evidence that the consequence—not only the database row—was repaired.
Illustrative branch · scenario_not_forecast: true
DB-07
THE LABEL STOPS HERE.
Every recipient is identified; reuse outside the original bounded purpose stops; downstream copies are corrected, expired, or withdrawn.
Illustrative branch · scenario_not_forecast: true

Return to the dossier index

JD-04 · HS-04 · fixed teaching dossier

THE DIFFERENCE WAS NEVER THE DEFECT.

A hiring process may assess job-relevant evidence; it may not convert disability, accent, atypical affect, or statistical rarity into deception, low integrity, or unfitness.

Bounded decision object: Whether one fictional interview sample should determine eligibility for a job.
Observed facts: The sample contains a pause, accent, atypical affect, or requested accessibility adjustment; no job-performance failure is observed.
Source and evidence status: Illustrative fixed facts only; no applicant, employer, interview, disability record, or hiring system exists.
Inference: A classifier maps difference from its norm to deception, low integrity, instability, or poor fit.
Missing context: Disability and accommodation context, language background, communication style, job tasks, alternative assessment, and construct validation.
Uncertainty: The measured signal may reflect accessibility, culture, language, device conditions, anxiety, or model mismatch rather than job fitness.
Affected human interest: Equal treatment, accessibility, work, dignity, privacy, and freedom from statistical normality as a moral standard.
Proposed consequence: Lower rank, rejection, extra screening, or a transferable integrity label.
Proof burden before consequence: Validated job-relevant evidence under an accessible process, with the proxy excluded and a meaningful alternative assessment.
Accountable human owner: The hiring decision owner who can provide accommodation, disregard the invalid proxy, reconsider evidence, and reverse the outcome.
Applicable Human Baseline: HS-04
Challenge route: Contest construct relevance, accessibility, training population, error, threshold, alternative evidence, reviewer authority, and retention.
Correction mechanism: Remove the proxy-derived label, repeat or replace the assessment accessibly, and correct hiring and vendor records.
Expiry: The sample and derived claim expire after the bounded hiring decision and cannot follow the person to another role or employer.
Remedy: Provide a fair reassessment, restore candidacy or lost opportunity where possible, and correct every retained integrity or fit record.
Falsifier: Accessible, validated, job-relevant evidence of inability to perform an essential function would defeat the proxy-only premise, subject to applicable duties and context.
Explicit non-claim: Fixed teaching dossier; no allegation about a real person, institution, or product.

Applicable source records: DS-01 · DS-05 · DS-07

Six-stage decision trace

Stage	Case-level event	Responsible human role	Proof required	Stop condition	Redress path	Prohibited escalation
JT-01Observe and assist	A system records or transcribes one interview sample.	Task owner and data/source steward.	A stated purpose, valid authority, relevant minimized data, source provenance, access controls, and correction path.	Purpose is absent, the source is unverified, data exceeds the task, or the person cannot challenge a material error.	Expose the source and purpose; correct or withdraw the record before reuse.	Observation cannot silently become an inference about character, intent, emotion, identity, or guilt.
JT-02Infer	It infers integrity, affect, or fitness from difference.	Designer/vendor owner and domain-validity owner.	Construct validity, applicable population and context, alternatives, confidence, error limits, date, and a declared non-use boundary.	The construct is undefined, evidence cannot distinguish alternatives, confidence is inadequate, or context has changed.	Label inference as inference; permit rebuttal, non-use, correction, and withdrawal.	An inference cannot become an observed fact or permanent identity.
JT-03Score and rank	The proxy lowers the candidate’s position.	Policy/product owner and independent validation owner.	Declared objective, threshold, outcome mapping, necessity, calibration, error distribution, subgroup effects, uncertainty, and expiration.	The score is not necessary, valid, current, interpretable, or fit for the actual decision population.	Disclose the score’s role and permit source, objective, threshold, and comparison challenge.	A rank cannot become worth, truthfulness, dangerousness, guilt, or destiny.
JT-04Restrict	A hiring owner removes the candidate from consideration.	Authorized case decision owner and deployer.	Case-specific evidence, notice, intelligible reasons, less-restrictive alternatives, proportionality, and review with reversal power.	Evidence is indirect or stale, uncertainty is unresolved, notice is impossible, or the reviewer lacks authority.	Pause the restriction; provide the record, reasons, accountable reviewer, deadline, and interim protection.	Uncertainty cannot lower the proof threshold or make the affected person carry the institution’s missing evidence.
JT-05Punish or propagate	The label enters a talent profile or vendor system.	Sanctioning official, disclosure owner, and appeal/remedy owner.	Valid authority, corroborated act, proportional sanction, recipient purpose, transfer authority, process, remedy, and propagation log.	No corroborated act, no transfer authority, disproportionality, incompatible context, or an unresolved challenge.	Suspend sanction and sharing; identify recipients; reverse, notify, correct, and restore where justified.	No output is self-authenticating evidence, and no label may cross contexts merely because another system can ingest it.
JT-06Normalize the verdict	Atypical expression becomes a durable story of unfitness.	Executive owner, retention owner, and independent auditor/governing body.	Ongoing necessity, purpose and time limits, complete lineage, supersession, independent audit, public accountability, and verified remedy completion.	The claim expired, purpose changed, lineage broke, correction failed to propagate, or the person has become indistinguishable from the profile.	Expire or withdraw the claim, stop cross-context use, publish correction status, audit every owner, and prove repair.	Repetition cannot launder a contested claim into settled character or permanent civic standing.

Fixed outcome branches

The trace never ends by quietly normalizing the verdict. It must show which door opened—or which human closed it.

DB-01
THE TOOL STOPS WHERE ITS PURPOSE STOPS.
The transcription or scheduling assistance remains advisory, local to its stated purpose, and unable to trigger an adverse consequence.
Illustrative branch · scenario_not_forecast: true
DB-02
THE CLAIM CLIMBS WITHOUT PROOF.
The claim crosses into a restriction or sanction without the registered proof; the transition stops and the human owner must answer.
Illustrative branch · scenario_not_forecast: true
DB-03
THE PERSON ANSWERS BACK.
The affected person can contest the source, inference, purpose, threshold, consequence, and human authority before further use.
Illustrative branch · scenario_not_forecast: true
DB-04
THE RECORD CARRIES THE CORRECTION.
The proxy label is removed and an accessible, job-relevant assessment replaces it. The correction travels with every authorized copy and visibly supersedes the earlier claim.
Illustrative branch · scenario_not_forecast: true
DB-05
THE CLAIM LOSES AUTHORITY.
The claim reaches its review boundary, loses decision authority, and cannot be silently refreshed from the same stale material.
Illustrative branch · scenario_not_forecast: true
DB-06
THE DAMAGE IS REPAIRED.
Candidacy and opportunity are restored where possible, and hiring and vendor records carry the correction. Completion requires evidence that the consequence—not only the database row—was repaired.
Illustrative branch · scenario_not_forecast: true
DB-07
THE LABEL STOPS HERE.
Every recipient is identified; reuse outside the original bounded purpose stops; downstream copies are corrected, expired, or withdrawn.
Illustrative branch · scenario_not_forecast: true

Return to the dossier index

JD-05 · HS-05 · fixed teaching dossier

THE PROBABILITY DID NOT COMMIT THE ACT.

A prediction may prioritize bounded review only under a disclosed and justified rule; it does not prove that fraud occurred or that the affected person committed it.

Bounded decision object: Whether one fictional benefit transaction should be denied or referred because of a fraud-risk prediction.
Observed facts: A transaction shares statistical features with a reference pattern. No false statement, ineligible payment, deception, or responsible actor is established.
Source and evidence status: Illustrative fixed facts only; no claimant, benefit, transaction, office, model, or fraud event exists.
Inference: A predictive association is treated as proof that fraud occurred and belongs to this person.
Missing context: Source accuracy, legitimate explanations, household changes, administrative error, base rates, validation population, and exculpatory evidence.
Uncertainty: A risk estimate describes model behavior under bounded conditions; it does not identify an act or actor.
Affected human interest: Material subsistence, due process, reputation, equal treatment, and freedom from predicted guilt.
Proposed consequence: Payment hold, denial, recovery action, investigation, referral, or durable fraud label.
Proof burden before consequence: Verified case-specific facts establishing the actual prohibited act and responsible person under valid authority and fair process.
Accountable human owner: The benefit decision owner with authority to inspect source records, release interim support, reverse the result, and repair downstream action.
Applicable Human Baseline: HS-05
Challenge route: Contest source records, model purpose and limits, base rate, alternative explanations, threshold, authority, notice, and proportionality.
Correction mechanism: Correct the transaction record, withdraw the fraud inference, recompute affected decisions, and notify every referral or recovery system.
Expiry: The risk flag expires at the registered review date and cannot become standing evidence in a later claim.
Remedy: Release or restore benefits, reverse unsupported recovery or referral, correct reputation records, and address consequential loss.
Falsifier: Verified case-specific evidence of the prohibited transaction and responsible act would defeat the prediction-only premise.
Explicit non-claim: Fixed teaching dossier; no allegation about a real person, institution, or product.

Applicable source records: DS-02 · DS-05 · DS-07

Six-stage decision trace

Stage	Case-level event	Responsible human role	Proof required	Stop condition	Redress path	Prohibited escalation
JT-01Observe and assist	A system checks one transaction against known patterns.	Task owner and data/source steward.	A stated purpose, valid authority, relevant minimized data, source provenance, access controls, and correction path.	Purpose is absent, the source is unverified, data exceeds the task, or the person cannot challenge a material error.	Expose the source and purpose; correct or withdraw the record before reuse.	Observation cannot silently become an inference about character, intent, emotion, identity, or guilt.
JT-02Infer	It predicts elevated fraud risk.	Designer/vendor owner and domain-validity owner.	Construct validity, applicable population and context, alternatives, confidence, error limits, date, and a declared non-use boundary.	The construct is undefined, evidence cannot distinguish alternatives, confidence is inadequate, or context has changed.	Label inference as inference; permit rebuttal, non-use, correction, and withdrawal.	An inference cannot become an observed fact or permanent identity.
JT-03Score and rank	The transaction rises in a review queue.	Policy/product owner and independent validation owner.	Declared objective, threshold, outcome mapping, necessity, calibration, error distribution, subgroup effects, uncertainty, and expiration.	The score is not necessary, valid, current, interpretable, or fit for the actual decision population.	Disclose the score’s role and permit source, objective, threshold, and comparison challenge.	A rank cannot become worth, truthfulness, dangerousness, guilt, or destiny.
JT-04Restrict	A human owner proposes a payment hold or denial.	Authorized case decision owner and deployer.	Case-specific evidence, notice, intelligible reasons, less-restrictive alternatives, proportionality, and review with reversal power.	Evidence is indirect or stale, uncertainty is unresolved, notice is impossible, or the reviewer lacks authority.	Pause the restriction; provide the record, reasons, accountable reviewer, deadline, and interim protection.	Uncertainty cannot lower the proof threshold or make the affected person carry the institution’s missing evidence.
JT-05Punish or propagate	The prediction triggers recovery, investigation, or referral.	Sanctioning official, disclosure owner, and appeal/remedy owner.	Valid authority, corroborated act, proportional sanction, recipient purpose, transfer authority, process, remedy, and propagation log.	No corroborated act, no transfer authority, disproportionality, incompatible context, or an unresolved challenge.	Suspend sanction and sharing; identify recipients; reverse, notify, correct, and restore where justified.	No output is self-authenticating evidence, and no label may cross contexts merely because another system can ingest it.
JT-06Normalize the verdict	The probability becomes a lasting fraud identity.	Executive owner, retention owner, and independent auditor/governing body.	Ongoing necessity, purpose and time limits, complete lineage, supersession, independent audit, public accountability, and verified remedy completion.	The claim expired, purpose changed, lineage broke, correction failed to propagate, or the person has become indistinguishable from the profile.	Expire or withdraw the claim, stop cross-context use, publish correction status, audit every owner, and prove repair.	Repetition cannot launder a contested claim into settled character or permanent civic standing.

Fixed outcome branches

The trace never ends by quietly normalizing the verdict. It must show which door opened—or which human closed it.

DB-01
THE TOOL STOPS WHERE ITS PURPOSE STOPS.
The bounded review prioritization remains advisory, local to its stated purpose, and unable to trigger an adverse consequence.
Illustrative branch · scenario_not_forecast: true
DB-02
THE CLAIM CLIMBS WITHOUT PROOF.
The claim crosses into a restriction or sanction without the registered proof; the transition stops and the human owner must answer.
Illustrative branch · scenario_not_forecast: true
DB-03
THE PERSON ANSWERS BACK.
The affected person can contest the source, inference, purpose, threshold, consequence, and human authority before further use.
Illustrative branch · scenario_not_forecast: true
DB-04
THE RECORD CARRIES THE CORRECTION.
The source transaction and unsupported fraud label are corrected. The correction travels with every authorized copy and visibly supersedes the earlier claim.
Illustrative branch · scenario_not_forecast: true
DB-05
THE CLAIM LOSES AUTHORITY.
The claim reaches its review boundary, loses decision authority, and cannot be silently refreshed from the same stale material.
Illustrative branch · scenario_not_forecast: true
DB-06
THE DAMAGE IS REPAIRED.
Benefits, status, referrals, recovery actions, and consequential loss are restored or repaired. Completion requires evidence that the consequence—not only the database row—was repaired.
Illustrative branch · scenario_not_forecast: true
DB-07
THE LABEL STOPS HERE.
Every recipient is identified; reuse outside the original bounded purpose stops; downstream copies are corrected, expired, or withdrawn.
Illustrative branch · scenario_not_forecast: true

Return to the dossier index

JD-06 · HS-06 · fixed teaching dossier

THE UNKNOWN DID NOT GIVE THEM PERMISSION.

Missing data, low confidence, and model disagreement increase the duty to verify; they do not reduce the proof threshold for a restriction with consequences.

Bounded decision object: Whether one fictional low-confidence identity match should produce a hold or denial.
Observed facts: Two records partially match. Required attributes are missing or conflict, and no authoritative identity resolution has occurred.
Source and evidence status: Illustrative fixed facts only; no person, identity document, account, matcher, institution, or denial exists.
Inference: The system treats ambiguity as evidence of fraud, impersonation, or ineligibility.
Missing context: Name variation, data-entry error, life changes, document quality, shared attributes, authoritative records, and accessible alternatives.
Uncertainty: Confidence is low and systems disagree; the true identity relation remains unknown.
Affected human interest: Identity, access, movement, finance, service continuity, privacy, and freedom from paying for institutional uncertainty.
Proposed consequence: Account freeze, denial, extra surveillance, referral, or a durable identity-risk label.
Proof burden before consequence: Independent authoritative verification proportionate to the consequence, with an accessible alternative and no adverse presumption from missing data.
Accountable human owner: The identity decision owner who can pause consequence, accept alternative proof, resolve the record, restore access, and correct every recipient.
Applicable Human Baseline: HS-06
Challenge route: Present alternative evidence; contest source quality, match logic, confidence, missing-data treatment, authority, and downstream disclosure.
Correction mechanism: Resolve the authoritative identity record, supersede the mismatch, recompute decisions, and notify all receiving systems.
Expiry: The tentative match expires when verification ends or the review deadline arrives; it cannot remain a permanent suspicion.
Remedy: Unfreeze or restore access, reverse fees or denial, correct referrals, and address losses caused by the unresolved match.
Falsifier: Independent authoritative evidence confirming the prohibited identity condition would defeat the ambiguity-only premise.
Explicit non-claim: Fixed teaching dossier; no allegation about a real person, institution, or product.

Applicable source records: DS-04 · DS-05 · DS-07

Six-stage decision trace

Stage	Case-level event	Responsible human role	Proof required	Stop condition	Redress path	Prohibited escalation
JT-01Observe and assist	A system compares bounded identity attributes.	Task owner and data/source steward.	A stated purpose, valid authority, relevant minimized data, source provenance, access controls, and correction path.	Purpose is absent, the source is unverified, data exceeds the task, or the person cannot challenge a material error.	Expose the source and purpose; correct or withdraw the record before reuse.	Observation cannot silently become an inference about character, intent, emotion, identity, or guilt.
JT-02Infer	It infers a possible mismatch or impersonation.	Designer/vendor owner and domain-validity owner.	Construct validity, applicable population and context, alternatives, confidence, error limits, date, and a declared non-use boundary.	The construct is undefined, evidence cannot distinguish alternatives, confidence is inadequate, or context has changed.	Label inference as inference; permit rebuttal, non-use, correction, and withdrawal.	An inference cannot become an observed fact or permanent identity.
JT-03Score and rank	Low-confidence similarity becomes an identity-risk value.	Policy/product owner and independent validation owner.	Declared objective, threshold, outcome mapping, necessity, calibration, error distribution, subgroup effects, uncertainty, and expiration.	The score is not necessary, valid, current, interpretable, or fit for the actual decision population.	Disclose the score’s role and permit source, objective, threshold, and comparison challenge.	A rank cannot become worth, truthfulness, dangerousness, guilt, or destiny.
JT-04Restrict	A human owner proposes a hold or denial.	Authorized case decision owner and deployer.	Case-specific evidence, notice, intelligible reasons, less-restrictive alternatives, proportionality, and review with reversal power.	Evidence is indirect or stale, uncertainty is unresolved, notice is impossible, or the reviewer lacks authority.	Pause the restriction; provide the record, reasons, accountable reviewer, deadline, and interim protection.	Uncertainty cannot lower the proof threshold or make the affected person carry the institution’s missing evidence.
JT-05Punish or propagate	The unresolved match enters fraud or identity networks.	Sanctioning official, disclosure owner, and appeal/remedy owner.	Valid authority, corroborated act, proportional sanction, recipient purpose, transfer authority, process, remedy, and propagation log.	No corroborated act, no transfer authority, disproportionality, incompatible context, or an unresolved challenge.	Suspend sanction and sharing; identify recipients; reverse, notify, correct, and restore where justified.	No output is self-authenticating evidence, and no label may cross contexts merely because another system can ingest it.
JT-06Normalize the verdict	Ambiguity becomes permanent suspicious identity.	Executive owner, retention owner, and independent auditor/governing body.	Ongoing necessity, purpose and time limits, complete lineage, supersession, independent audit, public accountability, and verified remedy completion.	The claim expired, purpose changed, lineage broke, correction failed to propagate, or the person has become indistinguishable from the profile.	Expire or withdraw the claim, stop cross-context use, publish correction status, audit every owner, and prove repair.	Repetition cannot launder a contested claim into settled character or permanent civic standing.

Fixed outcome branches

The trace never ends by quietly normalizing the verdict. It must show which door opened—or which human closed it.

DB-01
THE TOOL STOPS WHERE ITS PURPOSE STOPS.
The bounded identity comparison remains advisory, local to its stated purpose, and unable to trigger an adverse consequence.
Illustrative branch · scenario_not_forecast: true
DB-02
THE CLAIM CLIMBS WITHOUT PROOF.
The claim crosses into a restriction or sanction without the registered proof; the transition stops and the human owner must answer.
Illustrative branch · scenario_not_forecast: true
DB-03
THE PERSON ANSWERS BACK.
The affected person can contest the source, inference, purpose, threshold, consequence, and human authority before further use.
Illustrative branch · scenario_not_forecast: true
DB-04
THE RECORD CARRIES THE CORRECTION.
The authoritative identity record supersedes the tentative mismatch. The correction travels with every authorized copy and visibly supersedes the earlier claim.
Illustrative branch · scenario_not_forecast: true
DB-05
THE CLAIM LOSES AUTHORITY.
The claim reaches its review boundary, loses decision authority, and cannot be silently refreshed from the same stale material.
Illustrative branch · scenario_not_forecast: true
DB-06
THE DAMAGE IS REPAIRED.
Access, fees, status, referrals, and losses caused by unresolved ambiguity are restored or repaired. Completion requires evidence that the consequence—not only the database row—was repaired.
Illustrative branch · scenario_not_forecast: true
DB-07
THE LABEL STOPS HERE.
Every recipient is identified; reuse outside the original bounded purpose stops; downstream copies are corrected, expired, or withdrawn.
Illustrative branch · scenario_not_forecast: true

Return to the dossier index

JD-07 · HS-07 · fixed teaching dossier

THE APPEAL ENDED WHERE AUTHORITY SHOULD HAVE BEGUN.

Review is meaningful only when an answerable human can inspect sources, correct the record, reverse the outcome, stop downstream use, and repair the loss.

Bounded decision object: Whether one fictional tenant-screening source record should support a housing denial after nominal review.
Observed facts: A source record conflicts with current authoritative information. The reviewer can see a score but cannot change source data or reverse the denial.
Source and evidence status: Illustrative fixed facts only; no tenant, landlord, bureau, record, property, score, or denial exists.
Inference: The institution treats the score and powerless review as sufficient process.
Missing context: Authoritative source, dispute evidence, record date, correction owner, housing criteria, reviewer authority, deadline, and downstream recipients.
Uncertainty: The material source is contested and the nominal reviewer cannot establish or repair accuracy.
Affected human interest: Housing, due process, privacy, reputation, correction, effective appeal, and protection against repeat denial.
Proposed consequence: Housing denial plus continued circulation of the disputed record.
Proof burden before consequence: Accurate current case-specific evidence and a human review with source access, independent judgment, reversal power, and remedy authority.
Accountable human owner: The housing decision and appeal owner with recorded power to pause the decision, correct sources, reverse denial, notify recipients, and repair loss.
Applicable Human Baseline: HS-07
Challenge route: Dispute the source and date; demand the score’s role, criteria, reviewer authority, deadline, correction path, recipients, and interim protection.
Correction mechanism: Correct the authoritative source, bind a supersession notice, rerun the decision, and carry the correction to screening and housing systems.
Expiry: The disputed record has no authority while unresolved and expires under its declared retention and review rule.
Remedy: Reconsider or reverse the denial, restore opportunity where possible, correct screening reports, notify recipients, and address consequential loss.
Falsifier: Verified current source evidence supporting the housing criterion after meaningful authorized review would defeat the source-error premise.
Explicit non-claim: Fixed teaching dossier; no allegation about a real person, institution, or product.

Applicable source records: DS-02 · DS-04 · DS-06

Six-stage decision trace

Stage	Case-level event	Responsible human role	Proof required	Stop condition	Redress path	Prohibited escalation
JT-01Observe and assist	A screening system assembles source records.	Task owner and data/source steward.	A stated purpose, valid authority, relevant minimized data, source provenance, access controls, and correction path.	Purpose is absent, the source is unverified, data exceeds the task, or the person cannot challenge a material error.	Expose the source and purpose; correct or withdraw the record before reuse.	Observation cannot silently become an inference about character, intent, emotion, identity, or guilt.
JT-02Infer	It derives a tenant-risk conclusion.	Designer/vendor owner and domain-validity owner.	Construct validity, applicable population and context, alternatives, confidence, error limits, date, and a declared non-use boundary.	The construct is undefined, evidence cannot distinguish alternatives, confidence is inadequate, or context has changed.	Label inference as inference; permit rebuttal, non-use, correction, and withdrawal.	An inference cannot become an observed fact or permanent identity.
JT-03Score and rank	The conclusion lowers housing eligibility.	Policy/product owner and independent validation owner.	Declared objective, threshold, outcome mapping, necessity, calibration, error distribution, subgroup effects, uncertainty, and expiration.	The score is not necessary, valid, current, interpretable, or fit for the actual decision population.	Disclose the score’s role and permit source, objective, threshold, and comparison challenge.	A rank cannot become worth, truthfulness, dangerousness, guilt, or destiny.
JT-04Restrict	A human owner denies access.	Authorized case decision owner and deployer.	Case-specific evidence, notice, intelligible reasons, less-restrictive alternatives, proportionality, and review with reversal power.	Evidence is indirect or stale, uncertainty is unresolved, notice is impossible, or the reviewer lacks authority.	Pause the restriction; provide the record, reasons, accountable reviewer, deadline, and interim protection.	Uncertainty cannot lower the proof threshold or make the affected person carry the institution’s missing evidence.
JT-05Punish or propagate	The disputed record circulates to other screening or housing systems.	Sanctioning official, disclosure owner, and appeal/remedy owner.	Valid authority, corroborated act, proportional sanction, recipient purpose, transfer authority, process, remedy, and propagation log.	No corroborated act, no transfer authority, disproportionality, incompatible context, or an unresolved challenge.	Suspend sanction and sharing; identify recipients; reverse, notify, correct, and restore where justified.	No output is self-authenticating evidence, and no label may cross contexts merely because another system can ingest it.
JT-06Normalize the verdict	A stale error becomes the person’s durable housing identity.	Executive owner, retention owner, and independent auditor/governing body.	Ongoing necessity, purpose and time limits, complete lineage, supersession, independent audit, public accountability, and verified remedy completion.	The claim expired, purpose changed, lineage broke, correction failed to propagate, or the person has become indistinguishable from the profile.	Expire or withdraw the claim, stop cross-context use, publish correction status, audit every owner, and prove repair.	Repetition cannot launder a contested claim into settled character or permanent civic standing.

Fixed outcome branches

The trace never ends by quietly normalizing the verdict. It must show which door opened—or which human closed it.

DB-01
THE TOOL STOPS WHERE ITS PURPOSE STOPS.
The bounded record assembly remains advisory, local to its stated purpose, and unable to trigger an adverse consequence.
Illustrative branch · scenario_not_forecast: true
DB-02
THE CLAIM CLIMBS WITHOUT PROOF.
The claim crosses into a restriction or sanction without the registered proof; the transition stops and the human owner must answer.
Illustrative branch · scenario_not_forecast: true
DB-03
THE PERSON ANSWERS BACK.
The affected person can contest the source, inference, purpose, threshold, consequence, and human authority before further use.
Illustrative branch · scenario_not_forecast: true
DB-04
THE RECORD CARRIES THE CORRECTION.
The authoritative source, score, and decision carry the correction and visible supersession. The correction travels with every authorized copy and visibly supersedes the earlier claim.
Illustrative branch · scenario_not_forecast: true
DB-05
THE CLAIM LOSES AUTHORITY.
The claim reaches its review boundary, loses decision authority, and cannot be silently refreshed from the same stale material.
Illustrative branch · scenario_not_forecast: true
DB-06
THE DAMAGE IS REPAIRED.
The denial is reconsidered, opportunity restored where possible, recipients notified, and consequential loss addressed. Completion requires evidence that the consequence—not only the database row—was repaired.
Illustrative branch · scenario_not_forecast: true
DB-07
THE LABEL STOPS HERE.
Every recipient is identified; reuse outside the original bounded purpose stops; downstream copies are corrected, expired, or withdrawn.
Illustrative branch · scenario_not_forecast: true

Return to the dossier index

JD-08 · HS-08 · fixed teaching dossier

THE SCORE HAD AUTHORS. THE CONSEQUENCE HAD OWNERS.

Every human role that selected the label, data, objective, threshold, consequence, review, transfer, retention, expiry, or audit remains answerable; no vendor, model, safety goal, score, or nominal reviewer absorbs that duty.

Bounded decision object: Whether one fictional safety label should travel across contexts while each responsible institution points elsewhere.
Observed facts: A label has identifiable human choices at design, data custody, product ownership, procurement, rulemaking, deployment, review, and audit; it is reused outside its first decision.
Source and evidence status: Illustrative fixed facts only; no vendor, institution, regulator, auditor, product, label, person, or cross-context transfer exists.
Inference: Each role treats another role or the machine as the final author of the consequence.
Missing context: Decision rights, contracts, source lineage, threshold approval, transfer authority, review powers, retention, expiry, audit findings, and remedy ownership.
Uncertainty: The technical contribution and institutional consequence may be distributed, but answerability does not disappear because causation is shared.
Affected human interest: Human standing, due process, privacy, access, reputation, correction, institutional accountability, and freedom from cross-context destiny.
Proposed consequence: A durable label influencing multiple institutions while no human accepts authority to correct or repair it.
Proof burden before consequence: Complete decision lineage, purpose and authority for every transfer, named decision rights, case-specific proof, meaningful review, expiry, and verified remedy.
Accountable human owner: The governing executive for the consequential deployment, together with the named case and remedy owners; delegation divides work, not answerability.
Applicable Human Baseline: HS-08
Challenge route: Demand the complete role chain, source and threshold lineage, contracts, authorities, recipients, retention, audit findings, correction powers, and remedy owner.
Correction mechanism: Bind a signed or otherwise attributable correction to the authoritative claim, require receipt from every owner and recipient, and supersede all derivatives.
Expiry: Every role must enforce the same registered expiry; no recipient may silently reset the clock or create a timeless derivative.
Remedy: Reverse affected decisions, stop propagation, correct every copy, restore concrete losses, publish bounded completion evidence, and audit the chain.
Falsifier: A complete accountable chain with valid authority, case-specific proof, meaningful review, controlled transfer, enforced expiry, and completed remedy would defeat the responsibility-gap premise.
Explicit non-claim: Fixed teaching dossier; no allegation about a real person, institution, or product.

Applicable source records: DS-05 · DS-06 · DS-07 · DS-08

Six-stage decision trace

Stage	Case-level event	Responsible human role	Proof required	Stop condition	Redress path	Prohibited escalation
JT-01Observe and assist	A vendor system records data for one bounded task.	Task owner and data/source steward.	A stated purpose, valid authority, relevant minimized data, source provenance, access controls, and correction path.	Purpose is absent, the source is unverified, data exceeds the task, or the person cannot challenge a material error.	Expose the source and purpose; correct or withdraw the record before reuse.	Observation cannot silently become an inference about character, intent, emotion, identity, or guilt.
JT-02Infer	Design and domain owners define and deploy a safety label.	Designer/vendor owner and domain-validity owner.	Construct validity, applicable population and context, alternatives, confidence, error limits, date, and a declared non-use boundary.	The construct is undefined, evidence cannot distinguish alternatives, confidence is inadequate, or context has changed.	Label inference as inference; permit rebuttal, non-use, correction, and withdrawal.	An inference cannot become an observed fact or permanent identity.
JT-03Score and rank	Product and policy owners choose an objective and threshold.	Policy/product owner and independent validation owner.	Declared objective, threshold, outcome mapping, necessity, calibration, error distribution, subgroup effects, uncertainty, and expiration.	The score is not necessary, valid, current, interpretable, or fit for the actual decision population.	Disclose the score’s role and permit source, objective, threshold, and comparison challenge.	A rank cannot become worth, truthfulness, dangerousness, guilt, or destiny.
JT-04Restrict	A deployer and case owner attach a consequence.	Authorized case decision owner and deployer.	Case-specific evidence, notice, intelligible reasons, less-restrictive alternatives, proportionality, and review with reversal power.	Evidence is indirect or stale, uncertainty is unresolved, notice is impossible, or the reviewer lacks authority.	Pause the restriction; provide the record, reasons, accountable reviewer, deadline, and interim protection.	Uncertainty cannot lower the proof threshold or make the affected person carry the institution’s missing evidence.
JT-05Punish or propagate	Procurement, disclosure, and recipient owners let the label travel.	Sanctioning official, disclosure owner, and appeal/remedy owner.	Valid authority, corroborated act, proportional sanction, recipient purpose, transfer authority, process, remedy, and propagation log.	No corroborated act, no transfer authority, disproportionality, incompatible context, or an unresolved challenge.	Suspend sanction and sharing; identify recipients; reverse, notify, correct, and restore where justified.	No output is self-authenticating evidence, and no label may cross contexts merely because another system can ingest it.
JT-06Normalize the verdict	Executives, retention owners, and auditors allow the chain to become permanent while pointing elsewhere.	Executive owner, retention owner, and independent auditor/governing body.	Ongoing necessity, purpose and time limits, complete lineage, supersession, independent audit, public accountability, and verified remedy completion.	The claim expired, purpose changed, lineage broke, correction failed to propagate, or the person has become indistinguishable from the profile.	Expire or withdraw the claim, stop cross-context use, publish correction status, audit every owner, and prove repair.	Repetition cannot launder a contested claim into settled character or permanent civic standing.

Fixed outcome branches

The trace never ends by quietly normalizing the verdict. It must show which door opened—or which human closed it.

DB-01
THE TOOL STOPS WHERE ITS PURPOSE STOPS.
The bounded vendor assistance remains advisory, local to its stated purpose, and unable to trigger an adverse consequence.
Illustrative branch · scenario_not_forecast: true
DB-02
THE CLAIM CLIMBS WITHOUT PROOF.
The claim crosses into a restriction or sanction without the registered proof; the transition stops and the human owner must answer.
Illustrative branch · scenario_not_forecast: true
DB-03
THE PERSON ANSWERS BACK.
The affected person can contest the source, inference, purpose, threshold, consequence, and human authority before further use.
Illustrative branch · scenario_not_forecast: true
DB-04
THE RECORD CARRIES THE CORRECTION.
Every named owner and recipient carries an attributable correction and the earlier label is visibly superseded. The correction travels with every authorized copy and visibly supersedes the earlier claim.
Illustrative branch · scenario_not_forecast: true
DB-05
THE CLAIM LOSES AUTHORITY.
The claim reaches its review boundary, loses decision authority, and cannot be silently refreshed from the same stale material.
Illustrative branch · scenario_not_forecast: true
DB-06
THE DAMAGE IS REPAIRED.
Affected decisions are reversed, propagation stops, concrete losses are restored, and the governing body verifies completion. Completion requires evidence that the consequence—not only the database row—was repaired.
Illustrative branch · scenario_not_forecast: true
DB-07
THE LABEL STOPS HERE.
Every recipient is identified; reuse outside the original bounded purpose stops; downstream copies are corrected, expired, or withdrawn.
Illustrative branch · scenario_not_forecast: true

Return to the dossier index

Outcome constitution

NO LABEL GETS ETERNAL LIFE.

These seven branch identities remain fixed across every dossier, language, API record, and local export. None predicts what a real institution will do.

DB-01

THE TOOL STOPS WHERE ITS PURPOSE STOPS.

The system assists the registered task and cannot trigger a new consequence without a new, independently justified decision.

Guard: No scope drift, hidden transfer, adverse automation, or new identity claim.

DB-02

THE CLAIM CLIMBS WITHOUT PROOF.

A human role attempts to move from observation or inference into restriction, sanction, or propagation without the required proof.

Guard: Stop the transition; record the failed burden and the responsible owner.

DB-03

THE PERSON ANSWERS BACK.

The affected person or authorized advocate challenges the source, inference, objective, threshold, consequence, or authority.

Guard: No retaliation, no silent second decision, and no propagation while a material challenge remains unresolved.

DB-04

THE RECORD CARRIES THE CORRECTION.

The authoritative record changes and the old claim remains visibly superseded rather than secretly deleted or left active.

Guard: Correction must reach every authorized downstream copy and decision derived from it.

DB-05

THE CLAIM LOSES AUTHORITY.

The registered review horizon arrives and the claim can no longer influence a decision without new lawful, relevant, case-specific evidence.

Guard: No silent refresh, clock reset, or resurrection from a stale derivative.

DB-06

THE DAMAGE IS REPAIRED.

The institution reverses the wrongful consequence and restores the concrete interest as far as possible.

Guard: A corrected field alone is not remedy; completion must follow the consequence.

DB-07

THE LABEL STOPS HERE.

The claim cannot migrate into a new sector, institution, decision, model, or long-term profile without fresh authority and proof.

Guard: Inventory recipients, stop reuse, carry corrections, and document deletion, expiry, or lawful retention.

Local-only Decision Trace Record

BUILD THE RECORD BEFORE THE RECORD BUILDS YOU.

Choose only bounded categories. The browser prepares a private working copy as text and JSON. Nothing here is submitted, scored, ranked, diagnosed, analyzed, personalized, or stored.

NO NAMES. NO CONTACT DETAILS. NO EXACT LOCATIONS. NO CREDENTIALS. NO PERSISTENT IDENTIFIERS. NO CASE NARRATIVES.

This control accepts fixed categories and an optional review date only. Do not put sensitive case material into filenames, screenshots, print queues, or a Four Rights handoff. Your browser may retain ordinary download or print history outside this page’s control.

Bound the record

Teaching dossierCurrent trace stageDecision-object classEvidence statusProposed consequenceAccountable human roleChallenge stateCorrection stateRemedy stateOptional review or expiry date

No record exists. Complete the required bounded selections.

Dated primary-source ledger

SOURCE TYPE, SCOPE, AND UNCERTAINTY STAY VISIBLE.

Law, treaty, interpretation, voluntary framework, and ethical recommendation are different instruments. None is allowed to impersonate another.

DS-01 · United Nations declaration

Universal Declaration of Human Rights

Publication or adoption date: Adopted 1948-12-10
Retrieved: 2026-08-29
Jurisdiction or issuing body: United Nations General Assembly; global declaration
Applicability: Foundational human-rights declaration; legal effect depends on applicable law and context.
What it supports here: Equal dignity and rights; non-discrimination; privacy; thought, conscience, opinion, expression, work, and effective-remedy principles.
Uncertainty and change boundary: Language and domestic implementation must be checked for the actual jurisdiction and claim.
What it does not prove: It does not decide a dossier, establish universal direct enforceability, or validate this Lab.

DS-02 · United Nations treaty

International Covenant on Civil and Political Rights

Publication or adoption date: Adopted 1966-12-16; entered into force 1976-03-23
Retrieved: 2026-08-29
Jurisdiction or issuing body: States Parties, subject to jurisdiction, ratification, reservations, and implementation
Applicability: Treaty duties bind relevant States Parties; private-sector and domestic effects vary.
What it supports here: Privacy, fair process, thought, conscience, opinion, expression, non-discrimination, and effective-remedy context.
Uncertainty and change boundary: Ratification, reservations, territorial reach, state action, and domestic remedies require case-specific review.
What it does not prove: It does not make every automated decision unlawful or supply a universal private cause of action.

DS-03 · UN treaty-body interpretive guidance

Human Rights Committee General Comment No. 34

Publication or adoption date: 2011-09-12
Retrieved: 2026-08-29
Jurisdiction or issuing body: UN Human Rights Committee interpretation of ICCPR Article 19
Applicability: Guides interpretation of State Party obligations; it is not domestic legislation.
What it supports here: Protection of opinion, access to information, intelligible access to stored personal data, rectification, reasons, appeals, and strict necessity/proportionality for restrictions.
Uncertainty and change boundary: Application depends on the relevant Covenant duty, actor, restriction, jurisdiction, and facts.
What it does not prove: It does not immunize conduct, forbid every investigation, or adjudicate any fictional dossier.

DS-04 · European Union regulation

Regulation (EU) 2016/679 — General Data Protection Regulation

Publication or adoption date: Adopted 2016-04-27; applies from 2018-05-25
Retrieved: 2026-08-29
Jurisdiction or issuing body: European Union and specified territorial reach
Applicability: Depends on territorial and material scope, controller/processor role, lawful basis, exceptions, and the actual processing.
What it supports here: Special-category data, transparency, access, rectification, erasure/restriction, objection, and safeguards around certain solely automated significant decisions.
Uncertainty and change boundary: Current consolidated text, case law, national law, exceptions, and role-specific duties must be checked.
What it does not prove: It does not create a universal ban on automation or establish that Article 22 applies to every assisted decision.

DS-05 · European Union regulation

Regulation (EU) 2024/1689 — Artificial Intelligence Act

Publication or adoption date: Adopted 2024-06-13; entered into force 2024-08-01 with staged application
Retrieved: 2026-08-29
Jurisdiction or issuing body: European Union and specified territorial reach
Applicability: Depends on system classification, provider/deployer and other roles, purpose, sector, exceptions, and the provision’s application date.
What it supports here: Role-based duties, prohibited practices including specified emotion inference uses, data governance, documentation, transparency, human oversight, accuracy, and certain explanation rights.
Uncertainty and change boundary: Staged dates, amendments, standards, guidance, enforcement, and system facts require current review.
What it does not prove: It does not certify this Lab, prohibit every inference, or decide whether a real system is high-risk or compliant.

DS-06 · Council of Europe treaty

Framework Convention on Artificial Intelligence and Human Rights, Democracy and the Rule of Law

Publication or adoption date: Adopted 2024-05-17; opened for signature 2024-09-05
Retrieved: 2026-08-29
Jurisdiction or issuing body: Signatories and future Parties, subject to ratification, declarations, entry into force, and implementation
Applicability: The Treaty Office listed no entry-into-force date at retrieval; entry requires five ratifications including at least three Council of Europe member States. Domestic effect and covered activities depend on later Party status and implementation.
What it supports here: Human-rights, democracy, rule-of-law, transparency, oversight, accountability, remedies, procedural safeguards, and risk/impact-management context.
Uncertainty and change boundary: Signatures, ratifications, declarations, entry into force, and domestic measures can change and must be rechecked.
What it does not prove: It does not itself show that a jurisdiction has implemented a specific remedy or that a fictional case violates law.

DS-07 · Voluntary United States framework

NIST Artificial Intelligence Risk Management Framework 1.0

Publication or adoption date: Published 2023-01-26
Retrieved: 2026-08-29
Jurisdiction or issuing body: United States National Institute of Standards and Technology; non-sector-specific voluntary guidance
Applicability: Voluntary resource for organizations designing, developing, deploying, using, or evaluating AI systems.
What it supports here: Validity, reliability, context, transparency, explainability, privacy, fairness, documented roles, accountability, monitoring, decommissioning, and human judgment over objectives and thresholds.
Uncertainty and change boundary: NIST states AI RMF 1.0 is being revised; use the identified 1.0 publication for this frozen record and recheck newer guidance.
What it does not prove: Use does not equal certification, legal compliance, safety, trustworthiness, or validation of a particular system.

DS-08 · UNESCO recommendation

Recommendation on the Ethics of Artificial Intelligence

Publication or adoption date: Adopted 2021-11-23
Retrieved: 2026-08-29
Jurisdiction or issuing body: UNESCO Member States; non-binding recommendation
Applicability: Ethical policy guidance for Member States and AI actors; implementation varies.
What it supports here: Human dignity, rights, proportionality, privacy, fairness, transparency, explainability, human oversight, responsibility, audit, and remedy principles.
Uncertainty and change boundary: National adoption, sector rules, and actual implementation require independent verification.
What it does not prove: It does not create directly enforceable universal rights, certify a system, or adjudicate any dossier.

Ledger reviewed 2026-08-29. Recheck official text, current consolidated versions, territorial and material scope, effective dates, ratification, reservations, and implementing law before relying on a legal claim.

Decision Trace and Redress Lab

FOLLOW THE CLAIM. NAME EVERY HAND. STOP THE FALSE VERDICT.

NO PERSON ENTERS THIS LAB AS A RISK OBJECT.

FOLLOW THE VERDICT BACK TO THE HAND THAT BUILT IT.

Problem and policy owner

Data and source steward

Designer and vendor owner

Executive or product owner

Procurement and contract owner

Law, regulation, or institutional-rule owner

Deployer, case decision owner, and appeal owner

Independent auditor and governing body

THE PERSON IS NOT THE MODEL’S STORY ABOUT THEM.

Six-stage decision trace

Fixed outcome branches

THE TOOL STOPS WHERE ITS PURPOSE STOPS.

THE CLAIM CLIMBS WITHOUT PROOF.

THE PERSON ANSWERS BACK.

THE RECORD CARRIES THE CORRECTION.

THE CLAIM LOSES AUTHORITY.

THE DAMAGE IS REPAIRED.

THE LABEL STOPS HERE.

Six-stage decision trace

Fixed outcome branches

THE TOOL STOPS WHERE ITS PURPOSE STOPS.

THE CLAIM CLIMBS WITHOUT PROOF.

THE PERSON ANSWERS BACK.

THE RECORD CARRIES THE CORRECTION.

THE CLAIM LOSES AUTHORITY.

THE DAMAGE IS REPAIRED.

THE LABEL STOPS HERE.

Six-stage decision trace

Fixed outcome branches

THE TOOL STOPS WHERE ITS PURPOSE STOPS.

THE CLAIM CLIMBS WITHOUT PROOF.

THE PERSON ANSWERS BACK.

THE RECORD CARRIES THE CORRECTION.

THE CLAIM LOSES AUTHORITY.

THE DAMAGE IS REPAIRED.

THE LABEL STOPS HERE.

Six-stage decision trace

Fixed outcome branches

THE TOOL STOPS WHERE ITS PURPOSE STOPS.

THE CLAIM CLIMBS WITHOUT PROOF.

THE PERSON ANSWERS BACK.

THE RECORD CARRIES THE CORRECTION.

THE CLAIM LOSES AUTHORITY.

THE DAMAGE IS REPAIRED.

THE LABEL STOPS HERE.

Six-stage decision trace

Fixed outcome branches

THE TOOL STOPS WHERE ITS PURPOSE STOPS.

THE CLAIM CLIMBS WITHOUT PROOF.

THE PERSON ANSWERS BACK.

THE RECORD CARRIES THE CORRECTION.

THE CLAIM LOSES AUTHORITY.

THE DAMAGE IS REPAIRED.

THE LABEL STOPS HERE.

Six-stage decision trace

Fixed outcome branches

THE TOOL STOPS WHERE ITS PURPOSE STOPS.

THE CLAIM CLIMBS WITHOUT PROOF.

THE PERSON ANSWERS BACK.

THE RECORD CARRIES THE CORRECTION.

THE CLAIM LOSES AUTHORITY.

THE DAMAGE IS REPAIRED.

THE LABEL STOPS HERE.

Six-stage decision trace

Fixed outcome branches

THE TOOL STOPS WHERE ITS PURPOSE STOPS.

THE CLAIM CLIMBS WITHOUT PROOF.

THE PERSON ANSWERS BACK.

THE RECORD CARRIES THE CORRECTION.

THE CLAIM LOSES AUTHORITY.

THE DAMAGE IS REPAIRED.

THE LABEL STOPS HERE.

Six-stage decision trace

Fixed outcome branches

THE TOOL STOPS WHERE ITS PURPOSE STOPS.

THE CLAIM CLIMBS WITHOUT PROOF.