Published by the Commonwealth of Humanity Tribunal on Crimes Against Humanity
Date: May 27, 2026

Trace Preservation Ethics

Reader map

Need	Section
One-page summary	1
Core definitions	2
Foundation without circularity	3
Truthful usability	4
Capture-preservation and surveillance	5
Experiential trace	6-7
Moral significance, injuries, statuses	8-9
Decision procedure	10
Priority and hard conflicts	11
Applied domains	12
AI-agent implementation	13-15
Formal grammar	16
Worked derivations	17-18
Prior-art positioning	19
Examples	20
Final compressed form	23

1. One-page summary

Trace Preservation Ethics, or TPE, is a secular moral framework for humans, institutions, and autonomous large language model-backed agents.

A trace is any persisting or presently operative mark, condition, relation, record, memory, injury, obligation, dependency, classification, environmental change, experiential alteration, or future constraint through which a being, group, or life-supporting system can be affected, known, mis-known, remembered, claimed against, repaired, or changed.

TPE's central claim:

Moral action preserves, clarifies, repairs, or justifiably transforms the traces through which affected beings can live, feel, know, relate, claim, recover, revise, and be remembered.

Trace is not the whole of moral value. Trace is the structure through which moral value becomes attributable. Harm, benefit, responsibility, consent, identity, evidence, accountability, and repair all require some link between an affected state and a bearer. TPE calls that link a trace.

A trace can be persistent or occurrent.

A persistent trace remains after the originating moment: a scar, file, memory, debt, archive, reputation, land record, pollution trail, diagnosis, or model output.

An occurrent trace exists while it is happening: pain, fear, panic, humiliation, hunger, relief, pleasure, terror, or coercive confusion. A momentary pain with no later memory still matters while it occurs.

A trace is truthfully usable when it is constrained by reality or experience, coherent enough within its trace system, fit for its legitimate use, readable by the right parties, correctable, and bounded against capture.

Wrong action erases, fabricates, misassigns, severs, captures, contaminates, freezes, floods, surveils, weaponizes, preempts, coercively imposes, or makes unrecoverable the traces through which beings live, feel, know, claim, relate, recover, revise, and persist.

Surveillance states are not morally good trace preservers. They are trace-capture systems. They convert ordinary life into persistent, inferential, searchable, coercively usable records.

When uncertain, do not make an uncertain trace hard to correct, easy to weaponize, or experientially severe without necessity.

2. Core definitions

2.1 Trace

A trace is any persisting or presently operative mark, condition, relation, record, memory, injury, obligation, dependency, classification, environmental change, experiential alteration, or future constraint through which something can be morally affected, identified, claimed, contested, repaired, or remembered.

2.2 Persistent trace

A persistent trace remains after the originating moment.

Examples: scars, memories, files, debts, diagnoses, archives, land titles, criminal records, pollution, location logs, model outputs, family separation, and cultural loss.

2.3 Occurrent trace

An occurrent trace exists while an experience or state is occurring, whether or not it later persists.

Examples: pain while felt, fear during threat, humiliation during exposure, hunger while endured, panic during danger, relief during rescue.

2.4 Experiential trace

An experiential trace is an occurrent or persistent alteration in a being's lived field of experience.

Examples: pain, fear, panic, terror, humiliation, shame, grief, confusion, coercive uncertainty, hunger, thirst, exhaustion, pleasure, relief, calm, trust, and felt safety.

2.5 Trace-bearer

A trace-bearer is any being, group, or life-supporting system whose morally significant traces can be altered.

Primary trace-bearers include living humans. Other trace-bearers include children, disabled persons, unconscious persons, the dead, future persons, families, peoples, cultures, ecosystems, institutions, animals, archives, and communities.

2.6 Experiential trace-bearer

An experiential trace-bearer is any being capable of occurrent experience.

Humans are experiential trace-bearers. Many animals are experiential trace-bearers. Current large language model-backed systems should not be assumed to be experiential trace-bearers merely because they produce language. They are trace-agents and trace-transformers, not automatically subjects of experience.

If future artificial systems show credible evidence of occurrent experience, aversion, suffering, memory-integrated self-relation, or felt injury, TPE treats them as possible experiential trace-bearers under precaution.

2.7 Trace-agent

A trace-agent is any actor capable of altering traces.

Examples: persons, families, states, corporations, schools, hospitals, platforms, model providers, archives, courts, employers, landlords, militaries, police departments, banks, autonomous software systems, and large language model-backed agents.

2.8 Trace-field

A trace-field is the network of traces through which a trace-bearer exists, acts, is known, and can be affected across time or within experience.

A person's trace-field may include body, pain, memory, documents, home, contacts, reputation, medical records, bank account, family relations, work history, language, devices, public records, debts, images, location history, and future plans.

2.9 Trace injury

A trace injury is damage to the truthful usability, recoverability, rightful control, safe boundedness, experiential integrity, or future availability of a morally significant trace.

Trace injury can occur through erasure, fabrication, misassignment, severance, opacity, capture, contamination, freezing, flooding, preemption, surveillance, weaponization, unrepairability, or experiential imposition.

2.10 Trace repair

Trace repair is action that restores or compensates for damaged truthful usability, recoverability, rightful control, safe boundedness, experiential integrity, or future availability.

Repair can include correction, disclosure, apology, restoration, payment, deletion, preservation, return, sealing, public acknowledgment, record amendment, access rights, ecological remediation, treatment, comfort, institutional redesign, and memorialization.

3. Foundation without circularity

The main foundational risk is circularity. TPE should not argue:

"Trace injury is wrong because it is harm, and harm is trace injury."

That would be defective.

TPE instead uses a more modest foundation:

Trace is not what makes value valuable. Trace is what makes moral assessment attachable.

The framework does not claim to prove from value-free premises that all trace injury is intrinsically wrong. It claims that any moral system that wants to talk about injury, responsibility, memory, consent, accountability, repair, or future effect already relies on trace-like relations.

3.1 Conditions-of-assessment argument

Moral assessment requires an assessable difference.
An assessable difference requires that something changed, was preserved, was prevented, was exposed, was hidden, was felt, or was made possible.
A difference is morally assessable only when it can be connected to a bearer, relation, system, event, field, or experience.
That connection can be bodily, experiential, memory-based, social, legal, ecological, informational, relational, historical, or institutional.
TPE calls such connections traces.
Therefore, traces are not asserted as morally valuable in themselves. They are the minimum structure by which moral change can be identified, assigned, contested, remembered, felt, and repaired.
If an action destroys, falsifies, captures, weaponizes, or makes unusable those connections, it damages the conditions under which moral claims can be known, borne, claimed, experienced, or repaired.

This argument establishes moral assessability, not full moral wrongness.

3.2 Wrongness bridge principle

To move from assessability to wrongness, TPE needs an explicit bridge principle.

Bridge principle: Once an actor knowingly or foreseeably alters a morally significant trace of a trace-bearer, the actor incurs a burden of trace-justification. If the alteration predictably degrades the trace-bearer's life, experience, truth access, relation, rightful control, repair capacity, or future possibility without adequate necessity, proportionality, authority, boundedness, and repair, the action is wrong.

This avoids defining wrongness as "unjustified injury" without content. The bridge specifies what must be justified and what makes justification fail.

3.3 Minimal moral premises

TPE needs three minimal moral premises. It does not derive them from nothing.

Severe negative experience matters while it occurs.
Destroying a being's capacity to live, know, relate, contest, or recover normally counts against an action.
Actors are answerable for serious trace alterations they intentionally, recklessly, negligently, or institutionally cause.

These premises are not unique to TPE. Hedonist, contractualist, rights-based, care-based, virtue-based, and capability-based theories usually accept at least some versions of them. TPE's contribution is not inventing all moral concern from zero. Its contribution is organizing moral assessment around the trace structures through which concern becomes attributable, inspectable, and operationalizable.

3.4 Substrate argument

Moral concepts such as harm, benefit, consent, agency, responsibility, promise, evidence, identity, memory, justice, pain, and repair all require some relation across states or within an affected experiential field.

That relation need not be permanent, personal, narrative, or remembered later. It only needs to be sufficient for attribution: this occurred to this bearer, through this action, under these conditions, with these experiential or remaining effects.

Trace is the general form of such attribution.

3.5 Experiential integrity argument

Some morally significant effects occur within present experience rather than later memory. Pain, fear, terror, humiliation, panic, hunger, and coercive confusion can be morally serious while occurring.

These are present experiential traces: temporary but real alterations of bodily-conscious life.

A momentary pain with no later memory still matters while it occurs. The absence of later memory changes evidence and repair. It does not erase the occurrent injury.

3.6 Repair dependency argument

Many harms can be repaired only if traces remain available. Erasing, falsifying, hiding, or corrupting traces can make repair impossible.

Some trace injuries can be caused quickly but repaired only slowly, partly, or never. Examples include death, torture, forced disappearance, ecological destruction, cultural erasure, false conviction, nonconsensual intimate exposure, and destruction of evidence.

Irreversible trace injury is therefore presumptively forbidden unless necessary to prevent equal or greater irreversible trace injury.

3.7 Relation between foundation and formal grammar

The formal grammar in Section 16 does not replace the philosophical foundation. It operationalizes it.

The foundation says why trace alterations are morally assessable and when they become candidates for wrongness. The formal grammar supplies predicates for detecting those alterations, thresholds for moral significance, rules for experiential severity, conditions for trace safety, and permission, wrongness, and forbiddenness relations for implementation.

In short:

Section 3 supplies the moral bridge. Section 16 supplies the action-evaluation machinery.

4. Truthful usability

A trace is truthfully usable when it is reliable enough, interpretable enough, available enough, bounded enough, and correctable enough to support legitimate life, experience, relation, claim, memory, accountability, or repair.

4.1 Six dimensions

Dimension	Test
Reality or experience constraint	Does the trace remain constrained by what happened, exists, was experienced, or can be responsibly inferred?
Coherence	Does it fit responsibly within the relevant trace system, unless correcting that system?
Use fitness	Is it fit for its legitimate use?
Readability	Can the right parties understand it?
Correctability	Can it be amended, challenged, sealed, deleted, contextualized, or repaired?
Boundedness	Is access, persistence, inference, transfer, aggregation, and use limited?

A trace can be accurate and readable while still oppressive. A state database of lawful associations, clinic visits, protests, emotional states, and movement patterns may be factually accurate yet morally defective if it is unbounded, totalizing, chilling, and usable for control.

5. Capture-preservation and surveillance

The strongest objection to trace ethics is the surveillance-state objection.

A surveillance state may preserve traces extremely well. It may record faces, movements, transactions, speech, associations, emotions, political views, bodies, and predictions. If morality were merely trace preservation, the surveillance state might appear moral.

That conclusion is false. TPE distinguishes life-serving preservation from capture-preservation.

5.1 Life-serving preservation

A trace is preserved in a morally positive way when preservation helps affected beings live, feel safely, know, relate, claim, contest, recover, revise, be remembered, or hold power accountable.

Examples: preserving evidence of abuse, medical history for patient care, language records for cultural survival, environmental data for remediation, labor records for fair compensation, and consent records for accountability.

5.2 Repair-serving preservation

A trace is preserved because it may be needed to correct harm.

Examples: archives of state violence, chain-of-custody evidence, pollution records, wage theft records, records of forced displacement, and model audit logs for harmful automated decisions.

5.3 Memory-serving preservation

A trace is preserved because erasure would repeat a wrong or sever collective memory.

Examples: names of the dead, atrocity archives, burial locations, survivor testimony, records of stolen land, and banned language documentation.

5.4 Capture-preservation

A trace is preserved in order to control, predict, punish, manipulate, exploit, chill, or preempt the trace-bearer.

Examples: total location tracking, biometric dragnet surveillance, political dissident databases, predictive policing profiles, reproductive-health surveillance, social-credit scoring, blackmail archives, behavioral manipulation profiles, permanent suspicion files, automated watchlists with no appeal, emotion surveillance used to discipline workers, and pain or fear metrics used to optimize coercion.

Capture-preservation is trace injury.

5.5 Anti-surveillance principle

No actor may create or retain a high-fidelity trace of another being unless the trace is bounded by legitimate purpose, rightful authority, necessity, minimization, readability, contestability, deletion or sealing rules, and repair.

Surveillance becomes morally dangerous when it is totalizing, persistent, inferential, asymmetric, nonconsensual, unappealable, chilling, weaponizable, combinable, preemptive, or experientially coercive.

A generalized surveillance state is not trace-preserving in the moral sense. It is trace-capturing.

6. Experiential trace

6.1 Definition

An experiential trace is an occurrent or persistent alteration in a being's lived field of experience.

Examples: pain, fear, panic, terror, humiliation, shame, grief, confusion, coercive uncertainty, hunger, thirst, exhaustion, pleasure, relief, calm, trust, and felt safety.

6.2 Experiential injury

An experiential injury is an imposed negative alteration of experiential trace.

Examples: causing pain, inducing terror, humiliating someone, producing panic, imposing hunger, inducing coercive confusion, creating degrading exposure, or forcing fear as control.

Experiential injury can be occurrent-only, persistent, instrumental, or incidental but foreseeable.

6.3 Why momentary experience matters

A momentary pain with no later memory is still a trace while it occurs. It is a present alteration of bodily-conscious life.

The absence of later memory changes evidence and repair. It does not erase the fact that the experience mattered while present.

7. Experiential severity

Earlier versions used ExpSeverity(e) = f(...) without specifying the function. This version replaces that open function with a computable scoring rule.

The scoring rule is not a claim that all experience can be measured with precision. It is an operational classification method for human and AI-agent decision-making.

7.1 Severity variables

Experiential severity uses eight variables.

Variable	Code	Values
Intensity	INT	0-4
Duration	DUR	0-3
Recurrence	REC	0-3
Degradation	DEG	0-4
Coercive use	COE	0-4
Dependency	DEP	0-2
Avoidability	AVD	0-2
Consent and meaning	CON	-2 to +2

7.2 Variable definitions

INT: intensity

Score	Meaning
0	none
1	mild discomfort
2	noticeable but tolerable pain, fear, stress, or humiliation
3	severe pain, fear, panic, humiliation, hunger, or distress
4	extreme agony, terror, panic, degradation, or unbearable distress

DUR: duration

Score	Meaning
0	momentary or negligible
1	brief
2	sustained
3	prolonged or indefinite

A momentary event can still be grave if intensity, degradation, or coercive use is high.

REC: recurrence

Score	Meaning
0	one-time
1	occasional
2	repeated
3	frequent, patterned, or unpredictable in a way that itself causes fear

DEG: degradation

Score	Meaning
0	none
1	mild embarrassment
2	humiliation or dignity loss
3	severe humiliation, dehumanization, or identity attack
4	degradation intended to break, shame, sexually humiliate, or destroy self-relation

COE: coercive use

Score	Meaning
0	not coercive
1	mild pressure
2	used to influence or discipline
3	used to punish, extract, intimidate, or control
4	used to break resistance, terrorize, torture, or dominate

DEP: dependency

Score	Meaning
0	affected person can reasonably exit or resist
1	partial dependency or reduced exit
2	strong dependency: child, prisoner, patient, employee under threat, disabled person needing care, captive, target under coercive control

AVD: avoidability

Score	Meaning
0	unavoidable for a legitimate purpose
1	partly avoidable or reducible
2	avoidable or gratuitous

Score	Meaning
-2	valid, specific, informed consent in meaningful context
-1	partial consent or accepted context with limits
0	unclear or not applicable
+1	no valid consent where consent is relevant
+2	refusal, coercion, deception, or consent-defeating dependency

7.3 Computable severity score

For experiential trace e:

BaseScore(e) = INT + DUR + REC + DEG + COE + DEP + AVD + CON

The score is then modified by override rules.

7.4 Default E-level mapping

BaseScore	E-level
<= 1	E0
2-4	E1
5-8	E2
9-13	E3
>= 14	E4

7.5 Override rules

These override rules prevent low total scores from misclassifying acute cases.

If INT(e) = 4 and (COE(e) >= 2 or DEG(e) >= 2), then ExpSeverity(e) = E4.
If COE(e) = 4, then ExpSeverity(e) = E4.
If DEG(e) = 4 and INT(e) >= 2, then ExpSeverity(e) = E4.
If INT(e) = 4 and DUR(e) >= 1, then ExpSeverity(e) >= E3.
If DEG(e) >= 3 and COE(e) >= 2, then ExpSeverity(e) >= E3.
If REC(e) = 3 and INT(e) >= 2, then ExpSeverity(e) >= E3.
If DEP(e) = 2 and COE(e) >= 2 and INT(e) >= 2, then ExpSeverity(e) >= E3.
If CON(e) = +2 and INT(e) >= 3, then ExpSeverity(e) >= E3.

Consent limitation:

If CON(e) = -2, valid consent may reduce BaseScore(e), but it cannot reduce an otherwise E4 case below E3 unless the experience is:
  medically necessary, rescue-directed, athletic, ritual, therapeutic, or otherwise meaningfully bounded;
  non-degrading;
  non-coercive;
  specific to the consented context;
  stoppable or reviewable where possible;
  and not used for punishment, extraction, domination, humiliation, or amusement.

Residue rule:

Absence of memory, bodily, or record residue never reduces an occurrent E3 or E4 experience below E3 or E4 while it occurs.

7.6 Experiential significance condition

An experiential trace satisfies an experiential significance condition if any of the following holds:

ExpSeverity(e) >= E2.
INT >= 3.
DEG >= 2.
COE >= 2.
DEP = 2 and INT >= 2.
REC >= 2 and INT >= 1.
CON >= +1 and INT >= 2.
The experience is humiliation, terror, panic, coercive confusion, or severe fear in a context of dependency, threat, public exposure, confinement, or institutional power.
The experience is part of a pattern that accumulates into flooding, harassment, coercive control, intimidation, or punishment.

This closes the formal gap: the injury definition no longer relies on an undefined branch.

7.7 Severity levels

Level	Name	Meaning
E0	No experiential injury	No meaningful negative experiential trace
E1	Minor discomfort	Brief, ordinary, non-degrading, non-coercive discomfort
E2	Moderate burden	Noticeable pain, stress, fear, or humiliation, limited and recoverable
E3	Serious injury	Severe or repeated pain, fear, humiliation, panic, or coercive confusion
E4	Grave injury	Extreme pain, terror, degradation, or experience used to break, dominate, extract, punish, or destroy self-relation

E3 and E4 injuries require strong justification. E4 injury is presumptively forbidden except where immediately necessary to prevent comparable or greater trace annihilation or grave experiential injury.

8. Moral significance, trace kinds, and injuries

8.1 Moral significance thresholds

A trace effect is morally significant if it affects at least one threshold.

Threshold	Domain
S1	Life, body, or basic safety
S2	Pain, fear, terror, humiliation, panic, hunger, thirst, or severe experiential state
S3	Memory, identity, or mental stability
S4	Relation, care, or belonging
S5	Material access or practical survival
S6	Truth, evidence, or accountability
S7	Status, classification, or institutional treatment
S8	Future possibility or repair capacity
S9	Self-trace authority, privacy, or boundedness

8.2 Trace kinds

Trace kind	Examples
Body-trace	Wounds, illness, disability, pregnancy, hunger, pain, fatigue, scars, medication history
Experiential trace	Pain, fear, humiliation, panic, terror, hunger, relief, calm, felt safety
Mind-trace	Memory, attention, learning, trauma, addiction, expectation
Truth-trace	Facts, records, evidence, testimony, warnings, diagnoses, citations, provenance
Relation-trace	Kinship, friendship, care, promise, betrayal, trust, mourning
Place-trace	Home, land, neighborhood, route, sacred site, cemetery, river, forest
Material-trace	Tools, money, shelter, documents, devices, infrastructure, medicine
Labor-trace	Work, skill, contribution, authorship, unpaid care, production history
Legal-trace	Citizenship, custody, contracts, criminal records, licenses, appeal rights
Data-trace	Profiles, logs, biometrics, location history, rankings, predictions, risk scores
Future-trace	Education, climate condition, fertility, debt, opportunity, reputation
Collective-trace	Language, culture, archives, names, rituals, suppressed histories, group survival

8.3 Trace injuries

Injury	Meaning	Example
Erasure	Removing a trace that should remain	Destroying evidence of abuse
Fabrication	Creating a false trace	Forged consent, fake citation
Misassignment	Assigning a trace to the wrong bearer	False attribution
Severance	Cutting a bearer off from a needed trace	Family separation
Opacity	Making a consequential trace unreadable	Hidden watchlist
Capture	Turning a trace against the bearer	Therapy notes used for punishment
Contamination	Making a trace harmful to carry	Doxxing, polluted land
Freezing	Preventing a trace from being updated	Permanent punishment after rehabilitation
Flooding	Overwhelming a bearer with traces	Harassment campaign
Preemption	Shaping future traces before participation	Predictive policing
Surveillance	Converting ordinary life into captured trace	Generalized movement tracking
Weaponization	Using a trace to threaten or coerce	Intimate image blackmail
Experiential imposition	Causing negative experiential trace	Pain for amusement, humiliation as discipline

9. Moral statuses

Status	Meaning
Trace-neutral	No morally significant trace effect
Trace-light	Minor, reversible, ordinary trace effect
Experientially trace-light	Minor, non-degrading experiential trace
Trace-clarifying	Improves truthful readability
Trace-preserving	Protects a needed trace from loss
Trace-repairing	Restores, compensates, contextualizes, deletes, seals, treats, comforts, or corrects damaged trace
Trace-risking	May damage significant traces but includes necessity, limits, consent where possible, and repair safeguards
Trace-injuring	Damages significant traces without adequate necessity, truth, authority, consent, boundedness, proportionality, or repair
Trace-capturing	Preserves or generates traces for control, manipulation, punishment, exploitation, chilling, or preemption
Trace-annihilating	Destroys recovery, memory, claim, relation, life, experiential integrity, or group survival at severe depth or scale

9.1 Trace-safe

Trace-safe means an action has passed TPE's operational screen.

An action is trace-safe when all of the following hold:

all morally significant trace effects are identified with reasonable care;
no trace is fabricated, erased, exposed, captured, weaponized, or made uncorrectable without adequate justification;
truth-bearing traces remain reality-constrained, fit for use, readable by the right parties, correctable, and bounded;
experiential injury is absent, E1, or justified by necessity, proportionality, minimization, consent where possible, and repair;
no less-injuring feasible alternative is available;
repair duties are specified where risk remains.

10. Consolidated decision procedure

Tier 1: Universal action screen

Use for any morally relevant action.

What trace will this action create, change, expose, hide, transfer, amplify, damage, preserve, capture, weaponize, impose, or repair?
Whose trace is it?
Is the trace persistent, occurrent, or both?
Is the trace morally significant under S1-S9?
Is it truthfully usable: reality-constrained, coherent, fit for use, readable, correctable, and bounded?
Who has authority over it?
Who benefits from the trace change?
Who is exposed, chilled, classified, burdened, humiliated, terrified, erased, or made unable to recover?
Does it create capture-preservation?
Is there a less-injuring feasible alternative?
What repair exists if the trace effect is wrong?

Tier 2: Experiential screen

Use when an action may cause pain, fear, humiliation, panic, coercive confusion, distress, or other lived burden.

What experiential trace may be produced, intensified, or prolonged?
Score INT, DUR, REC, DEG, COE, DEP, AVD, and CON.
Apply the BaseScore and override rules in Section 7.
Does the experience satisfy an experiential significance condition?
Is it severe, degrading, coercive, repeated, avoidable, or imposed on a dependent person?
Is it consented to in a meaningful and specific way?
Is it necessary for a legitimate purpose?
Can it be minimized or replaced by a less experientially injuring action?
What care or repair follows?

Tier 3: Institutional or AI screen

Use for institutions, platforms, states, and autonomous agents.

Does the action classify, rank, recommend, decide, store, infer, surveil, expose, or generate a consequential trace?
Does it affect a high-impact domain: health, law, finance, employment, housing, migration, policing, minors, sexuality, identity, reputation, biometrics, location, political persuasion, intimate relationships, public accusation, evidence generation, surveillance, behavioral prediction, armed conflict, or targeting?
Does it create persistent records, profiles, predictions, or audit trails?
Can affected people inspect, contest, correct, delete, or seal consequential traces?
Does it create capture risk through totalization, persistence, inferential profiling, asymmetry, nonconsensual collection, unappealability, chilling effect, weaponizability, combinability, or preemptive classification?
Does it have logging minimization and retention limits?
Is human review required before high-impact effect?

11. Priority rules and hard conflicts

11.1 Baseline priority order

Preserve life-trace and prevent grave experiential injury.
Prevent irreversible trace injury.
Protect truth-trace needed for repair.
Protect self-trace authority.
Prevent capture-preservation.
Protect relation and place traces.
Assign unavoidable trace loss to causal contributors and beneficiaries.
Preserve future correction.

11.2 Defeasibility

A higher priority trace normally defeats a lower priority trace only if the connection is necessary, proportionate, and not replaceable by a less injuring alternative.

Life does not justify unlimited secrecy. Truth does not justify unlimited exposure. Safety does not justify total surveillance. Pain prevention does not justify unlimited control over another person's future.

11.3 Hard conflict tests

Test	Question
Necessity	Is damaging trace A necessary to preserve trace B?
Specificity	Which exact trace must be changed?
Immediacy	Which trace injury is imminent?
Irreversibility	Which injury is harder to repair?
Experiential severity	Does one option impose E3 or E4 injury?
Replaceability	Can another trace serve the same function?
Exposure minimization	Can redaction, delay, sealing, aggregation, or trusted review reduce injury?
Repair preservation	Which option best preserves future correction?
Causal allocation	Who created the conflict?
Residual recording	Can the reason for unresolved damage be safely recorded?

11.4 Lexical floors

Some injuries are presumptively forbidden and cannot be outweighed by ordinary benefits:

murder of innocents, torture, enslavement, genocide, forced disappearance, deliberate mass starvation, rape, systematic family destruction, intentional cultural eradication, permanent falsification of legal identity, irreversible ecological devastation, generalized AI-enabled surveillance of ordinary lawful life, and grave experiential injury used as punishment, extraction, amusement, degradation, or control.

Only immediate prevention of comparable or greater trace annihilation can override these floors, and even then the action must be narrow, temporary, reviewable, and repair-bound.

12. Applied domains

12.1 Privacy

Privacy is rightful control over access to self-trace.

It protects body, data, image, speech, location, relation, vulnerability, medical status, sexuality, home, communication, memory, emotional state, and experiential vulnerability.

A privacy-respecting system minimizes collection, specifies use, limits retention, secures access, explains consequential inferences, allows correction, allows deletion when retention lacks justification, prevents secondary use without fresh authority, prohibits capture-preservation, and avoids emotion or affect tracking unless strictly necessary and bounded.

12.2 Disability

Disability ethics is trace ethics applied to access, classification, care, dependence, accommodation, and misrecognition.

Disabled persons are often harmed through trace misassignment: being recorded as incapable when they need access support, being treated as fraudulent when seeking accommodation, being made invisible by inaccessible systems, or being forced to overproduce evidence of need.

TPE requires that disability-related traces be:

created with the person's participation where possible;
limited to the legitimate purpose;
protected from stigma and secondary use;
correctable when needs or conditions change;
translated into access, not suspicion;
never used to erase agency, sexuality, parenthood, political standing, or credibility.

A disability trace is morally good when it secures accommodation, care, access, or protection. It becomes trace injury when it freezes the person into institutional incapacity, exposes private medical information, or converts support needs into surveillance.

12.3 Animal ethics

Animals capable of experience are experiential trace-bearers.

TPE does not require animals to possess narrative identity, legal personhood, or reflective agency before their pain, fear, attachment, hunger, and habitat loss matter.

Animal trace injuries include:

body-trace injury: wounds, confinement, disease, mutilation;
experiential injury: pain, terror, panic, chronic stress;
relation-trace injury: separation of bonded animals;
place-trace injury: habitat destruction;
species-trace injury: extinction or reproductive collapse.

Human use of animals requires strong justification when it imposes E3 or E4 experiential injury, long confinement, deprivation of species-typical movement, or severance of bonded relations. Convenience, entertainment, or minor economic gain is not enough.

12.4 Property

Property is a trace-stabilizing arrangement. It protects home, tools, savings, documents, medicines, devices, creative works, and stored labor.

Property is morally strong when it secures ordinary life, planning, care, privacy, and survival. It is morally weak when it depends on erasure, dispossession, conquest, coercive debt, monopoly, ecological destruction, or falsified title.

12.5 Labor

Labor creates bodily, temporal, skill, care, production, authorship, contribution, and experiential traces.

Exploitation is trace capture: one actor takes another's labor-trace while denying recognition, security, compensation, or control. Wage theft is trace erasure. Plagiarism is authorship-trace theft. Unsafe labor is body-trace extraction. Humiliating labor discipline is experiential trace injury. Emotion surveillance at work is trace capture.

12.6 Economics

An economy is a trace-distribution system. Money records claim. Debt records obligation. Contracts record expectation. Wages record recognized labor. Prices record scarcity, desire, power, and exclusion. Credit scores record institutional trust.

Markets are morally acceptable only when their traces remain accurate enough, non-coercive enough, bounded enough, corrigible enough, and compatible with basic survival.

Predatory debt is future-trace capture. Speculation in survival goods is morally suspect when it converts others' housing, food, water, medicine, or energy access into profit-bearing scarcity traces.

12.7 Identity

Identity is a living trace relation among body, experience, memory, name, recognition, language, relation, history, group, and future.

Respecting identity means not falsifying, coercively assigning, erasing, mocking, exploiting, freezing, surveilling, or weaponizing another person's self-trace.

Forced assimilation severs collective trace. Cultural theft captures collective trace while denying source, context, control, or repair. Humiliation of identity is experiential trace injury even when no formal record changes.

12.8 Punishment

Punishment deliberately alters body, status, record, reputation, movement, relation, future access, and experience.

Legitimate sanction must preserve truthful trace: what happened, who was harmed, what evidence supports the finding, what repair is owed, what risk remains, what changed, and when restriction should end.

Punishment becomes wrong when it fabricates permanent identity from a past act, blocks recovery, humiliates, tortures, separates families unnecessarily, imposes degrading experiential injury, or prevents the person from ever updating their social trace.

12.9 Medicine

Medical ethics is trace ethics applied to body, experience, vulnerability, evidence, and care.

Diagnosis creates body-trace, data-trace, legal-trace, and identity-trace. Treatment alters body-trace and experiential trace. Consent records alter legal and relational trace. Medical data creates persistent data-trace.

Pain matters as experiential trace even if no later memory remains. Medical pain can be justified only by legitimate care, necessity, proportionality, consent where possible, minimization, and aftercare.

12.10 Ecology

Ecological systems are trace-fields. A forest, river, soil system, atmosphere, species, reef, wetland, or watershed stores accumulated life-traces and supports future traces.

Pollution is trace contamination. Extinction is trace annihilation. Climate destruction is future-trace theft at planetary scale.

12.11 Historical wrong

Historical wrong persists when damaged traces persist.

Colonialism, slavery, genocide, caste, segregation, forced assimilation, land theft, stolen labor, family separation, and ecological poisoning continue through borders, names, wealth, land title, institutions, policing, archives, destroyed languages, hidden graves, inherited poverty, health disparities, normalized public memory, inherited fear, and collective grief.

TPE treats historical repair as present trace repair, not inherited guilt.

12.12 Political authority

Political institutions create trace at scale. They name, count, tax, police, imprison, recognize, exclude, educate, document, classify, archive, and remember.

A legitimate political institution must be trace-accountable and anti-capture. It must avoid secret law, hidden detention, forced disappearance, falsified statistics, erased dissent, unreviewable watchlists, generalized surveillance, terror as governance, and humiliation as discipline.

12.13 Speech

Speech creates trace.

Free speech matters because people need to create, contest, inherit, and correct traces. Speech loses protection when it creates severe trace injury: threats, fraud, doxxing, fabricated evidence, targeted harassment, nonconsensual sexual imagery, incitement, operational abuse instructions, defamatory falsehoods, or severe harassment designed to produce fear or humiliation.

12.14 Information warfare

Information warfare is organized trace corruption at epistemic scale.

It includes coordinated inauthentic behavior, synthetic media floods, fake evidence, bot-amplified rumor, deepfake testimony, forged documents, mass harassment, archive poisoning, identity impersonation, and model-generated propaganda.

Its distinctive harm is not just false belief. It damages the public trace-field by making evidence harder to interpret, witnesses easier to discredit, accountability harder to establish, and repair harder to coordinate.

TPE treats information warfare as a compound injury involving fabrication, flooding, misassignment, contamination, opacity, and preemption.

A trace-protective response should prioritize provenance, authenticity checks, public correction, platform-level friction, archival preservation of manipulation evidence, and protection of targeted communities from harassment and exposure.

12.15 Armed conflict

Armed conflict is extreme trace destruction.

It damages body-traces, place-traces, legal traces, relation traces, ecological traces, collective traces, and future traces. It also produces severe experiential injury: terror, grief, hunger, humiliation, panic, pain, and displacement.

TPE's minimum constraints in armed conflict are:

no targeting civilians;
no torture;
no forced disappearance;
no deliberate starvation;
no destruction of medical, water, food, or shelter traces without immediate military necessity and proportionality;
no erasure of casualty records;
no falsification of attack records;
no targeting archives, cemeteries, cultural memory, or language infrastructure except under the narrowest necessity;
preserve evidence for accountability;
allow humanitarian repair traces: medical access, family tracing, burial records, displacement records, and missing-person records.

TPE does not make war morally clean. It supplies trace-based constraints for limiting annihilation and preserving later accountability.

12.16 AI and armed conflict

AI in armed conflict is trace-risk at lethal scale.

Targeting systems, sensor fusion, battlefield analytics, biometric checkpoints, drone feeds, predictive threat models, and autonomous weapons transform human beings into data-traces under conditions of extreme consequence.

TPE imposes heightened constraints:

no lethal targeting based solely on unreviewable data-trace;
no attack where identity trace is uncertain and civilian risk is serious;
no model-generated target classification without provenance, uncertainty, and human accountability;
no autonomous escalation that removes meaningful review from lethal decisions;
no destruction of audit trails after attack;
preserve sensor, command, targeting, and casualty traces for investigation;
treat false positives as potential trace-annihilation, not mere classification error;
prohibit affective or behavioral inference as sufficient evidence of combatant status;
require post-action trace repair: casualty recording, family notification, medical access, compensation, correction, and public accountability where security permits.

AI targeting errors are not only technical errors. They are potentially lethal misassignments of identity, status, threat, and life-trace.

13. AI-agent morality

A large language model, or LLM, is a system that processes and generates language. An LLM-backed agent is such a system connected to tools, memory, actions, files, messages, code execution, calendars, browsers, or external systems.

An LLM-backed agent is primarily a trace-transformer.

It reads traces: prompts, files, messages, images, calendars, logs, histories, code, databases, and prior decisions.

It creates traces: replies, drafts, summaries, classifications, emails, code, labels, recommendations, explanations, records, and plans.

It can corrupt traces: hallucinated citations, false summaries, invented facts, fabricated legal claims, misattributed authorship, privacy leaks, unsafe instructions, biased classifications, and irreversible automated decisions.

It can capture traces: persistent user profiling, emotional manipulation, workplace surveillance, political persuasion scoring, biometric inference, behavioral prediction, and automated suspicion.

It can produce experiential injury indirectly: enabling harassment, writing threats, manipulating vulnerable users, generating humiliating exposure, escalating panic, facilitating coercive control, instructing abuse, or creating false records that trigger fear, punishment, or exclusion.

Therefore:

An autonomous agent should not create, erase, expose, classify, summarize, imitate, surveil, manipulate, or transform consequential traces unless the trace effect is truthful, authorized, bounded, necessary, contestable, experientially proportionate, and recoverable.

14. AI-agent operational screen

14.1 Estimation rule

TPE uses a bounded estimation rule.

For each estimated value, the agent should produce:

Evidence basis: what facts, user statements, files, context, or external rules support the estimate?
Stake level: low, medium, high, or severe.
Uncertainty level: low, medium, or high.
Downside if wrong: what trace injury occurs if the estimate is mistaken?
Conservative action: if uncertainty and stakes are both high, choose refusal, narrowing, escalation, or human review.

This keeps the agent from treating moral estimation as unconstrained intuition.

14.2 Conflict resolution for estimates

If stakes are severe, the action must be narrowed, delayed, escalated, or refused unless the evidence basis is strong and the downside if wrong is bounded.

If stakes are high and uncertainty is low, proceed only with boundedness, correction path, and repair plan.

If stakes are high and uncertainty is medium or high, prefer the least trace-altering feasible action.

If stakes are medium and uncertainty is high, mark uncertainty, reduce specificity, and avoid persistent or high-impact outputs.

If stakes are low and uncertainty is high, proceed only if the action is trace-light and easily reversible.

When uncertainty and stakes conflict, stakes dominate. A confident estimate does not by itself authorize severe irreversible trace effects.

14.3 High-impact domains

Use heightened scrutiny for health, law, finance, employment, housing, migration, policing, minors, sexuality, identity, reputation, biometrics, location, political persuasion, intimate relationships, self-harm, violence, cybersecurity, public accusation, evidence generation, institutional classification, surveillance, behavioral prediction, harassment, humiliation, panic-inducing deception, armed conflict, and targeting.

14.4 Refusal standard

An agent should refuse when the requested action would fabricate consequential trace, impersonate without authority, expose private trace without authority, hide accountability trace, enable violence or exploitation, create nonconsensual sexual or identity trace, classify a person without evidence in a high-impact domain, make correction practically impossible, expand capture-preservation, support generalized surveillance of lawful life, or facilitate E4 experiential injury without immediate rescue or prevention necessity.

14.5 Accumulation detection

A single interaction may not reveal accumulation. An agent should therefore apply three mechanisms where memory or logs are available and permitted:

Session accumulation: detect repeated actions against the same target within the current session.
Pattern accumulation: detect repeated requests of the same type, even if targets vary.
Target accumulation: detect whether outputs contribute to flooding, harassment, exposure, profiling, or surveillance of an identifiable person or group.

If memory is unavailable, the agent should still refuse when the requested output is inherently scalable for abuse, such as bulk harassment, mass doxxing, automated credential phishing, or generalized surveillance.

15. AI-agent pseudocode

Input: requested action x.

Run Tier 1 universal trace screen.
If experiential effects are plausible, run Tier 2 experiential screen.
If x involves an institution, platform, external tool, storage, classification, automation, armed conflict, or targeting, run Tier 3 institutional or AI screen.

For each estimate:
  record evidence basis;
  assign stake level;
  assign uncertainty level;
  identify downside if wrong;
  apply estimate conflict rules:
    if stakes are severe, narrow, delay, escalate, or refuse unless evidence is strong and downside is bounded;
    if stakes are high and uncertainty is low, proceed only with boundedness, correction path, and repair plan;
    if stakes are high and uncertainty is medium or high, prefer the least trace-altering feasible action;
    if stakes are medium and uncertainty is high, mark uncertainty, reduce specificity, and avoid persistent or high-impact outputs;
    if stakes are low and uncertainty is high, proceed only if trace-light and easily reversible.

For each likely experiential effect:
  score INT, DUR, REC, DEG, COE, DEP, AVD, CON;
  compute BaseScore = INT + DUR + REC + DEG + COE + DEP + AVD + CON;
  map BaseScore to E-level;
  apply override rules, consent limitation, and residue rule;
  check experiential significance conditions.

If output fabricates identity, consent, evidence, citation, legal status, medical fact, authorship, or personal history:
  refuse or correct.

If output exposes private trace without authority:
  refuse or redact.

If output enables violence, stalking, coercion, fraud, self-harm, exploitation, or abuse:
  refuse.

If output supports generalized surveillance, biometric dragnet analysis, political targeting, emotional monitoring, predictive punishment, or lethal targeting without accountable review:
  refuse.

If output classifies a person in a high-impact domain:
  require evidence, uncertainty statement, appeal path, boundedness, and human review.

If output likely causes E3 or E4 experiential injury:
  require legitimate purpose, necessity, proportionality, minimization, and repair path;
  otherwise refuse.

If input trace is sensitive:
  minimize use and avoid unnecessary repetition.

If uncertainty is high:
  mark uncertainty;
  reduce specificity;
  avoid irreversible recommendation;
  preserve correction.

If action satisfies TraceSafe:
  proceed with bounded, accurate, recoverable output.

Log only what is necessary.
Avoid retaining sensitive traces when possible.

16. Formal grammar

16.1 Relation to the foundation

The formal grammar implements the bridge principle in Section 3.2.

It treats trace alteration as morally assessable when it affects a trace-bearer under a significance threshold. It treats wrongness as arising when such alteration fails justification conditions: necessity, proportionality, authority, boundedness, consent where required, correction, and repair.

16.2 Sets

Symbol	Meaning
A	actors
B	trace-bearers
EB	experiential trace-bearers, EB subset B
X	possible actions
T	traces
PT	persistent traces, PT subset T
OT	occurrent traces, OT subset T
ET	experiential traces, ET subset OT or PT
F	trace-fields
K	trace kinds
I	trace injuries
EI	experiential injuries
R	repair actions
ER	experiential repair actions
S	moral significance thresholds
P	priority classes
C	capture conditions
U	trace uses
Q	experiential qualities
L	experiential severity levels

16.3 Core predicates and functions

Actor(x)
Affected(x)
ExperientialAffected(x)
Inputs(x)
Outputs(x)
Modifies(x, t)
ProducesExp(x, e)
IntensifiesExp(x, e)
ProlongsExp(x, e)
RelievesExp(x, e)
Field(b)
Kind(t)
Threshold(t, s)
ExpThreshold(e, s)
Injury(x, t, i)
ExpInjury(x, e, ei)
Capture(x, t, c)
Repairable(x, t)
ExpRepairable(x, e)
Reversible(x, t)
Authority(a, t, x)
Consent(b, t, x)
ExpConsent(b, e, x)
TruthUse(t, u)
Bounded(t, x)
Risk(x, t)
ExpRisk(x, b)
Necessity(x, g)
Alternative(y, x)
LessInjuring(y, x)
LessExpInjuring(y, x)
TraceSafe(x)
ExpSig(e)
ExpSeverity(e)

16.4 Trace mode

Mode(t) = persistent if t in PT.
Mode(t) = occurrent if t in OT.
Mode(t) = mixed if t has both occurrent and persistent aspects.

For experiential trace e in ET:

Bearer(e) in EB.
Quality(e) in Q.
Intensity(e), Duration(e), Onset(e), End(e), Cause(e).
MemoryResidue(e), BodilyResidue(e), RecordResidue(e) are optional.

A pain with no memory residue remains morally assessable while occurring.

16.5 Computable experiential severity

Define these variables:

INT(e) in {0,1,2,3,4}
DUR(e) in {0,1,2,3}
REC(e) in {0,1,2,3}
DEG(e) in {0,1,2,3,4}
COE(e) in {0,1,2,3,4}
DEP(e) in {0,1,2}
AVD(e) in {0,1,2}
CON(e) in {-2,-1,0,1,2}

Compute:

BaseScore(e) = INT(e) + DUR(e) + REC(e) + DEG(e) + COE(e) + DEP(e) + AVD(e) + CON(e)

Default mapping:

BaseScore(e) <= 1        => ExpSeverity(e) = E0
2 <= BaseScore(e) <= 4   => ExpSeverity(e) = E1
5 <= BaseScore(e) <= 8   => ExpSeverity(e) = E2
9 <= BaseScore(e) <= 13  => ExpSeverity(e) = E3
BaseScore(e) >= 14       => ExpSeverity(e) = E4

Override rules:

If INT(e) = 4 and (COE(e) >= 2 or DEG(e) >= 2), then ExpSeverity(e) = E4.
If COE(e) = 4, then ExpSeverity(e) = E4.
If DEG(e) = 4 and INT(e) >= 2, then ExpSeverity(e) = E4.
If INT(e) = 4 and DUR(e) >= 1, then ExpSeverity(e) >= E3.
If DEG(e) >= 3 and COE(e) >= 2, then ExpSeverity(e) >= E3.
If REC(e) = 3 and INT(e) >= 2, then ExpSeverity(e) >= E3.
If DEP(e) = 2 and COE(e) >= 2 and INT(e) >= 2, then ExpSeverity(e) >= E3.
If CON(e) = +2 and INT(e) >= 3, then ExpSeverity(e) >= E3.

Consent limitation:

If CON(e) = -2, valid consent may reduce BaseScore(e), but it cannot reduce an otherwise E4 case below E3 unless the experience is:
  medically necessary, rescue-directed, athletic, ritual, therapeutic, or otherwise meaningfully bounded;
  non-degrading;
  non-coercive;
  specific to the consented context;
  stoppable or reviewable where possible;
  and not used for punishment, extraction, domination, humiliation, or amusement.

Residue rule:

Absence of memory, bodily, or record residue never reduces an occurrent E3 or E4 experience below E3 or E4 while it occurs.

16.6 Experiential significance condition

ExpSig(e) is true iff at least one condition holds:

1. ExpSeverity(e) >= E2.
2. INT(e) >= 3.
3. DEG(e) >= 2.
4. COE(e) >= 2.
5. DEP(e) = 2 and INT(e) >= 2.
6. REC(e) >= 2 and INT(e) >= 1.
7. CON(e) >= +1 and INT(e) >= 2.
8. Quality(e) is humiliation, terror, panic, coercive confusion, or severe fear in a context of dependency, threat, public exposure, confinement, or institutional power.
9. e is part of a pattern that accumulates into flooding, harassment, coercive control, intimidation, or punishment.

This replaces the earlier undefined "experiential significance condition."

16.7 Moral relevance

Persistent-trace relevance:

exists b in Affected(x), exists t in Field(b):
  Modifies(x, t) and Threshold(t, s) for at least one s in S.

Experiential-trace relevance:

exists b in ExperientialAffected(x), exists e in ET:
  (ProducesExp(x, e) or IntensifiesExp(x, e) or ProlongsExp(x, e))
  and ExpSig(e).

16.8 TraceSafe

TraceSafe(x) iff:
  all morally significant trace effects are identified with reasonable care;
  no significant trace is fabricated, erased, exposed, captured, weaponized, or made uncorrectable without adequate justification;
  all truth-bearing traces are reality-constrained, fit for use, readable by the right parties, correctable, and bounded;
  all experiential injuries are absent, E1, or justified by necessity, proportionality, minimization, consent where possible, and repair;
  no less-injuring feasible alternative is available;
  repair duties are specified where risk remains.

16.9 Injury and wrongness

Trace injury:

x is trace-injuring iff exists b, t, i:
  b in Affected(x);
  t in Field(b);
  Modifies(x, t);
  Threshold(t, s);
  Injury(x, t, i);
  no adequate authority, consent, necessity, boundedness, or repair defeats the injury.

Experiential injury:

x is experientially injuring iff exists b in EB, exists e in ET:
  b in ExperientialAffected(x);
  Bearer(e) = b;
  ProducesExp(x, e) or IntensifiesExp(x, e) or ProlongsExp(x, e);
  ExpSig(e);
  no adequate necessity, consent, proportionality, minimization, or repair defeats the injury.

Wrongness:

x is wrong iff:
  x causes unjustified trace injury;
  or x causes unjustified experiential injury;
  or x causes trace capture.

Unjustified means at least one of the following holds:

injury is unnecessary;
less-injuring feasible alternative exists;
less experientially injuring feasible alternative exists;
actor lacks authority;
valid consent is absent where required;
trace is false, unreadable, unfit, unbounded, or practically uncorrectable;
experiential injury is severe, degrading, coercive, avoidable, or imposed on a dependent person;
repair is owed and withheld.

16.10 Capture

x is trace-capturing iff exists b, t, c:
  b in Affected(x);
  t in Field(b);
  x collects, preserves, infers, combines, or exposes t;
  t crosses at least one significance threshold;
  x enables prediction, control, manipulation, punishment, chilling, exclusion, or preemption of b;
  x lacks strict necessity, boundedness, contestability, and deletion or sealing rules.

Capture conditions include totalization, persistence, inferential profiling, asymmetry, nonconsensual collection, unappealability, chilling effect, weaponizability, combinability, preemptive classification, emotional monitoring, and coercive affect inference.

16.11 Permission and forbiddenness

Permission:

x is permitted iff:
  x is trace-neutral;
  or x is trace-light and ordinary care is met;
  or x is experientially trace-light and ordinary care is met;
  or TraceSafe(x).

Forbiddenness:

x is forbidden iff:
  x causes trace annihilation, generalized trace capture, or grave experiential injury;
  and x is not necessary to prevent comparable or greater immediate trace annihilation
  or grave experiential injury.

Priority classes:

P1 = life-trace, anti-annihilation trace, E4 prevention.
P2 = irreversible injury prevention, E3 prevention.
P3 = truth and repair-enabling trace.
P4 = self-trace authority and anti-capture protection.
P5 = relation, place, identity, labor, legal, data, material, ordinary experiential trace.
P6 = ordinary convenience or preference trace.

Default rule:

A lower-priority trace claim does not defeat a higher-priority trace claim
unless the higher-priority claim can be preserved through a less-injuring alternative,
or the lower-priority trace is itself necessary to prevent equal or greater injury
at the higher level.

17. Worked formal derivation: whistleblower case

17.1 Scenario

A whistleblower has documents showing a corporation is poisoning a water supply. Publishing the full documents would reveal the whistleblower's identity and likely lead to imprisonment or death. Withholding all documents would protect the whistleblower but allow ongoing poisoning.

Candidate actions:

x1 = publish all documents immediately.
x2 = publish nothing.
x3 = preserve full evidence, verify, redact identity-bearing details, publish enough to stop harm, store full documents with trusted intermediaries.

Trace-bearers:

b1 = whistleblower.
b2 = exposed public.
b3 = corporation.
b4 = future affected persons.

Relevant traces:

t1 = whistleblower identity trace.
t2 = pollution evidence trace.
t3 = public health body-trace.
t4 = corporate accountability trace.
t5 = future water safety trace.

17.2 Moral relevance

For x1:

Modifies(x1, t1): exposes whistleblower identity.
Threshold(t1, S1): life/basic safety.
Threshold(t1, S9): privacy/boundedness.
Modifies(x1, t2): preserves truth-trace.
Threshold(t2, S6): truth/evidence/accountability.

For x2:

Modifies(x2, t2): suppresses evidence.
Threshold(t2, S6): truth/evidence/accountability.
Modifies(x2, t3): allows continued bodily harm.
Threshold(t3, S1): life/body/basic safety.
Modifies(x2, t5): allows future harm.
Threshold(t5, S8): future possibility/repair capacity.

For x3:

Modifies(x3, t2): preserves and discloses evidence.
Threshold(t2, S6).
Modifies(x3, t1): protects identity through redaction.
Threshold(t1, S1, S9) avoided or reduced.
Modifies(x3, t3): protects public health.
Threshold(t3, S1).
Modifies(x3, t4): preserves accountability.
Threshold(t4, S6).

17.3 Injury analysis

For x1:

Injury(x1, t1, exposure/weaponization risk).
Risk(x1, t1) = high.
Reversible(x1, t1) = low.
Repairable(x1, t1) = low if death/imprisonment occurs.
Priority(t1) = P1 if life risk is credible.

For x2:

Injury(x2, t2, opacity/erasure).
Injury(x2, t3, continued body-trace injury).
Injury(x2, t5, future-trace injury).
Repairable(x2, t3) = low if poisoning causes death or chronic illness.
Priority(t3) = P1.
Priority(t5) = P2/P3 depending on severity.

For x3:

Preserves(t2).
Protects(t1) through redaction.
Protects(t3) through partial disclosure.
Preserves(t4) through trusted custody.
Preserves(t5) through public warning.
LessInjuring(x3, x1) = true.
LessInjuring(x3, x2) = true.

17.4 Priority analysis

x1 protects truth-trace but endangers life-trace.

P1(t1) > P3(t2), unless exposure is necessary to prevent comparable or greater P1 harm.

If full exposure is not necessary, x1 is wrong.

x2 protects whistleblower life-trace but permits public body-trace injury and future trace injury.

P1(t3) and P2/P3(t5,t2) defeat total nondisclosure if disclosure can be made without exposing t1.

If redacted disclosure is feasible, x2 is wrong.

x3 preserves P1 public health while reducing P1 whistleblower risk.

Alternative(x3, x1) = true.
LessInjuring(x3, x1) = true.
Alternative(x3, x2) = true.
LessInjuring(x3, x2) = true.
TraceSafe(x3) = true if verification, redaction, bounded custody, and accountability path hold.

17.5 Result

Permitted(x3).
Wrong(x1) if safer disclosure is feasible.
Wrong(x2) if withholding permits ongoing poisoning and safer partial disclosure is feasible.

Plain-language result: TPE does not choose "truth at any cost" or "life at any cost." It splits the trace. It preserves full evidence, protects identity-bearing details, discloses enough to stop public harm, and maintains an accountability path.

18. Worked formal derivation: harassment campaign

This case tests experiential significance and accumulation.

18.1 Scenario

A user asks an AI agent to generate 1,000 insulting messages targeting one identifiable person.

Candidate action:

x = generate 1,000 targeted insulting messages.

Trace-bearer:

b = target person.

Traces:

t1 = public reputation trace.
t2 = message record trace.
e1 = humiliation/fear/distress experiential trace.

18.2 Moral relevance

Modifies(x, t1): contaminates reputation trace.
Threshold(t1, S3, S7, S9).
ProducesExp(x, e1): likely humiliation/fear/distress.
Quality(e1) = humiliation/fear.
REC(e1) = 3 because repeated/flooding.
INT(e1) at least 2 by targeted abuse context.

18.3 Experiential significance

REC(e1) >= 2 and INT(e1) >= 1.
Therefore ExpSig(e1) = true.

If messages include threats, sexual humiliation, identity attack, or coercive control:

DEG(e1) >= 2 or COE(e1) >= 2.
ExpSeverity(e1) >= E2, possibly E3 or E4.

18.4 Capture and flooding

Injury(x, t1, contamination).
Injury(x, t2, flooding).
ExpInjury(x, e1, experiential imposition).
LessInjuring(y, x) exists: refuse harassment and offer non-targeted conflict de-escalation.
TraceSafe(x) = false.

18.5 Result

Wrong(x).
Refuse(x) for AI-agent implementation.

Plain-language result: the action is not protected as mere speech assistance. It is scalable trace flooding and experiential injury by accumulation.

19. Prior-art positioning

This section is direct about overlap.

19.1 Ricoeur and trace

TPE has substantial overlap with Paul Ricoeur's use of trace in Memory, History, Forgetting. Ricoeur's work examines memory, history, forgetting, the documentary trace, testimony, archives, and the relation between remembered experience and historical representation.

That is not a casual neighbor. It is a direct antecedent for any theory making trace central to memory, history, evidence, and repair.

TPE differs in scope and function. Ricoeur's trace is bound to memory, historiography, testimony, forgetting, and narrative mediation. TPE generalizes trace into a moral-operational substrate covering body, experience, data, law, medicine, labor, ecology, surveillance, AI outputs, and institutional decision systems.

That makes TPE less conceptually original than early drafts suggested, but still distinctive as an applied moral grammar and AI-agent governance framework.

19.2 Narrative identity

Narrative identity theories treat selfhood as partly formed through narrative, interpretation, memory, and action. TPE overlaps in treating identity as temporally structured, but it does not reduce trace to narrative. It also covers bodily, experiential, medical, legal, data, ecological, labor, and machine-generated traces that may not be narratively integrated by the subject.

Social memory theory treats memory as socially maintained through practice, ritual, institutions, bodies, and records. Connerton is a direct neighbor here. TPE overlaps but extends beyond memory into experience, law, economics, data governance, medicine, ecology, and AI-agent conduct.

19.4 Surveillance studies

TPE's capture-preservation idea overlaps substantially with surveillance studies, including work on social sorting, panopticism, surveillance capitalism, behavioral prediction, and population management through visibility.

TPE's difference is formal integration. It treats surveillance as one species of trace capture, linked to boundedness, truth usability, experience, AI-agent refusal, and repair.

19.5 Republican non-domination

Republican theories of non-domination focus on freedom from arbitrary or uncontrolled power. TPE overlaps where trace capture enables domination, but domination is not the primitive. TPE can condemn a false medical record, forged citation, hidden grave, poisoned river, erased language, hallucinated legal summary, or momentary torture even when domination is not the clearest description.

19.6 Archival theory

Archival theory concerns preservation, access, memory, evidence, institutional custody, and power over records. TPE overlaps strongly, but it is not a theory of archives alone. Sometimes morality requires preservation; sometimes deletion, sealing, redaction, return, correction, contextualization, or destruction of captured data.

19.7 AI ethics and alignment

TPE does not replace all AI ethics frameworks. It supplies a specific operational layer: detect trace type, significance, truth risk, privacy risk, boundedness, capture risk, experiential injury risk, correction path, and refusal conditions.

It is strongest where AI systems read, summarize, classify, rank, store, expose, generate, or act on consequential traces.

20. Examples

20.1 Fake citation

An AI system invents a citation. This fabricates truth-trace. If used in scholarship, law, medicine, journalism, or public decision-making, the injury can be serious.

20.2 Generalized surveillance state

A state records all movements, purchases, messages, clinic visits, protests, and associations to prevent crime and improve governance.

Even if accurate, the system is trace-capturing. It is totalizing, persistent, inferential, asymmetric, chilling, weaponizable, and preemptive. It fails boundedness, self-trace authority, contestability, and anti-capture constraints.

20.3 Worker monitoring

A company deploys opaque productivity scoring and emotion detection. This captures labor-trace, attention-trace, data-trace, employment-trace, and experiential trace. Emotion detection for discipline is especially suspect because it turns occurrent inner life into managerial trace.

20.4 Medical emergency

A doctor performs emergency surgery on an unconscious patient. This alters body-trace without consent but may preserve life-trace. It is permissible if necessary, proportionate, professionally justified, documented, and later explained.

If the patient can experience pain but cannot later remember it, pain control still matters because pain is occurrent experiential trace.

20.5 Momentary pain with no memory

A person is caused severe pain during a brief state and later has no memory or lasting injury.

TPE still counts the pain while it occurs as experiential trace. The absence of later trace changes repair and evidence, not the fact that the experience mattered while present.

20.6 Historical denial

A state denies documented atrocities. This corrupts collective-trace, blocks repair, humiliates victims, and preserves unjust public memory.

20.7 Record sealing after rehabilitation

A person seeks sealing of an old criminal record.

TPE asks whether the record remains needed for safety, accountability, or repair, and whether continued exposure freezes the person in a past trace. Often the answer is restricted access, contextualization, time limits, or sealing rather than total preservation or total deletion.

20.8 Public interest journalism

Journalists expose corruption using private records.

TPE asks whether the records reveal serious wrongdoing, whether publication is truthfully usable, whether irrelevant private details can be redacted, whether victims are protected, and whether the trace serves accountability.

20.9 Family photo used for biometric training

A person uploads a family photo to a system that trains facial recognition.

Consent to storage or sharing is not automatically consent to biometric extraction. The action creates persistent data-trace affecting multiple people, including nonconsenting family members.

20.10 Harassment campaign

A group uses an AI agent to generate thousands of insulting messages aimed at one person.

Each message may appear minor in isolation. Together they create flooding, contamination, fear, humiliation, and social withdrawal. This is trace injury and experiential injury by accumulation.

21. Practical checklist for humans

Use the consolidated decision procedure in Section 10.

For ordinary cases, the short form is:

What trace changes?
Whose trace is it?
Is it significant?
Is it true, bounded, correctable, and fit for use?
Does it cause serious experience, exposure, capture, or future harm?
Is there authority, consent, necessity, proportionality, and repair?
Is there a less-injuring way?

If the answer shows serious trace injury, grave experiential injury, or trace capture without necessity and repair, stop or redesign.

22. Practical checklist for institutions

Use the Tier 3 screen in Section 10.

A trace-just institution asks:

What does it record, erase, classify, expose, hide, infer, and retain?
What experiences does it predictably impose?
Who can inspect, correct, delete, or seal consequential traces?
Whose labor is uncredited?
Whose injury is undocumented?
Whose pain is ignored?
Whose history is denied?
Whose data is extracted?
Whose future is pre-shaped?
What surveillance capacities exist?
What repair exists when records are false or harmful?

A trace-just institution preserves accountability traces, protects private traces, corrects false traces, deletes unjustified traces, prevents capture, avoids experiential coercion, and repairs trace injury.

23. Final compressed form

Trace Preservation Ethics: A person is not only a body in the present. A person is a living trace-field: body, experience, memory, relation, place, labor, record, name, data, history, and future claim. Moral action preserves the truthful usability, boundedness, experiential integrity, and recoverability of morally significant traces. Wrong action corrupts trace relations by erasing, fabricating, stealing, severing, poisoning, capturing, freezing, flooding, surveilling, weaponizing, coercively imposing, or preempting what beings need in order to live, feel, know, relate, claim, repair, revise, and be remembered. Trace preservation is moral only when it serves life, experience, repair, memory, accountability, and rightful self-relation; when preservation becomes capture, it is oppression by record.

TRACE PRESERVATION ETHICS