artifacts/standard-named

Rendered from markdown source. Open raw source on GitHub.

--- title: "Telic Projection Estimation" subtitle: "Field, Projection, Mirror, Uncertainty, Correction, and the Limits of Inference" artifact_date: "2026-07-14" artifact_type: "candidate-foundational-paper" domain: "TELIC-FIELDS" scope: "WORKING" lineage: "THE-TELIC-FIELD-PAPERS" status: "candidate" processing_tier: 4 source_role: "derived-conceptual-artifact" content_canon_status: "unset" publication_status: "unpublished" series_position: "F.4" historical_title:

• "Simplified Telic Field Calculation"

derived_from:

• "20260714__TELIC-FIELDS__PAPER__CANDIDATE__F-1-V0-2__telic-fields-first-principles-foundation.md"
• "20260714__TELIC-FIELDS__PAPER__CANDIDATE__F-2__dyadic-composition-of-telic-fields.md"
• "20260714__TELIC-FIELDS__PAPER__CANDIDATE__F-3__the-constitutional-self.md"
• "20260714__TELIC-FIELDS__REVIEW__WORKING__G-2__terminology-and-primary-source-adjudication.md"

research_companion:

• "20260714__TELIC-FIELDS__REVIEW__WORKING__G-5__projection-preference-inference-and-epistemic-boundaries.md"

provenance_note: > The origin session used "simplified telic field calculation." G.1 and G.2 concluded that "calculation" implied precision and direct access the framework does not possess. "Telic projection estimation" is therefore the preferred technical term. The historical phrase remains preserved in lineage. ---

Telic Projection Estimation

Abstract

No relation, institution, or machine acts directly upon another center's complete telic field. It acts upon a representation: a request, requirement, role, model, metric, inferred preference, bodily signal, policy, contract, narrative, or other compressed projection of what appears to matter.

This paper defines telic projection estimation as the scoped, uncertain, and corrigible process of constructing a usable representation of end-relative significance for a particular relation or action. It distinguishes the living source field from the projection made available downstream and from the receiver mirror constructed through interpretation. It treats direct statements, observed behavior, model inference, delegated representation, and bodily or low-verbal signals as different evidence classes rather than interchangeable routes to a hidden true objective.

The paper develops scope, provenance, authority, confidence, expiration, privacy, correction, revocation, and projection-integrity requirements. It distinguishes field breach, projection breach, mirror breach, authority breach, and projection-integrity breach. It proposes a minimal Telic Projection Record and a formal model designed to preserve uncertainty rather than conceal it.

The governing invariant is:

The projection is what downstream action can see. The field is what still bears the consequences.

---

1. Why estimation replaces calculation

The origin conversation used the phrase simplified telic field calculation.

The phrase captured something important: downstream action cannot carry an entire living field and therefore requires a simplification. Yet calculation implies that the correct inputs exist, the operation is defined, and the resulting value can be treated as the answer.

None of those conditions can be assumed.

A person's field may be:

• partly unarticulated;
• internally plural;
• changing;
• context dependent;
• embodied;
• strategically withheld;
• uncertain even to the person;
• shaped by relationships and outer systems;
• differently represented at different times.

An institution's field may be distributed across:

• mission statements;
• budgets;
• incentives;
• policies;
• informal practice;
• executive behavior;
• technical architecture;
• legal constraints;
• customer dependence;
• survival pressures.

A model cannot simply calculate what the person or institution truly wants.

It can estimate a projection adequate—or inadequate—for a particular scope.

Estimation preserves the possibility that the source remains more complex than the system's representation.

The technical change is therefore substantive:

simplified telic field calculation
→ telic projection estimation

The older term remains part of the witnessed lineage. The newer term governs current technical work.

---

2. Field, projection, and mirror

Let \(A\) be a center of standing with local telic field \(T_A(t)\).

For a relation with \(B\), under scope \(s\), \(A\) may provide or authorize a projection:

\[ \widehat{T}_{A \rightarrow B}^{\,s,t} \]

The receiver \(B\) interprets that projection through its own context, history, incentives, language, and limitations:

\[ M_{A \mid B}^{\,s,t} = \mathcal{I}_B \left( \widehat{T}_{A \rightarrow B}^{\,s,t}, H_B, C_B \right) \]

The objects remain distinct:

\[ T_A \neq \widehat{T}_{A \rightarrow B} \neq M_{A \mid B} \]

2.1 Field

The field is the living landscape that continues bearing consequences.

It includes more than what is expressed or inferred.

2.2 Projection

The projection is a scoped representation made available for a relation or action.

It may be direct, delegated, inferred, or mixed.

2.3 Mirror

The mirror is what a receiver believes the projection means.

It can diverge even when the projection was accurate.

This three-part distinction prevents a common collapse:

what I said
→ what you heard
→ what the system acts upon

These stages should not become one object merely because they occur quickly.

---

3. Projection is relational and scoped

A projection is not a portable essence of the source.

The statement:

I want privacy.

may mean different things in different scopes:

• do not disclose my name publicly;
• do not retain this conversation;
• do not use this for model training;
• do not expose this medical fact to my employer;
• do not contact my family;
• let me participate pseudonymously;
• preserve the record but restrict access;
• let the witness know while the counterparty does not.

A projection without scope invites overreach.

The minimum scope dimensions include:

• recipient;
• purpose;
• action;
• affected domain;
• duration;
• expected consequence;
• downstream use;
• data handling;
• delegation;
• review trigger;
• exit.

The same source field can legitimately produce different projections for:

• a friend;
• a clinician;
• an employer;
• a court;
• a civic forum;
• a model;
• a public archive.

Difference across scopes does not automatically indicate deception.

It may indicate boundary competence.

A projection is adequate only relative to the action it is being asked to govern.

---

4. Evidence classes

Not every input to projection estimation has the same epistemic status.

4.1 Direct source statement

The center states what matters.

Example:

I do not consent to secondary use of this recording.

This is strong evidence of the source's declared projection for the stated scope.

It does not prove that the source has no conflicting concerns.

4.2 Source-confirmed reformulation

Another participant or model proposes a formulation, and the source confirms it.

Example:

Model: “Your boundary appears to be no use beyond this mediation.” Source: “Yes. That is correct.”

This becomes source-authorized but retains derivation provenance.

4.3 Delegated representation

An authorized person or process represents a center.

Examples include:

• legal guardian;
• corporate officer;
• elected representative;
• designated advocate;
• advance directive;
• pool governance rule.

Delegation must preserve scope and authority.

The delegate is not identical to the represented center.

4.4 Observed behavior

A system infers orientation from action.

Observed behavior may reveal operative tendencies, but the inference is underdetermined.

A person may remain in a job because they value it, fear poverty, lack alternatives, protect family, or feel trapped.

Behavior is evidence.

It is not transparent preference.

4.5 Historical pattern

Past actions or statements may predict current orientation.

They may also be stale.

History requires recency and context checks.

4.6 Bodily or low-verbal signal

The source may communicate through:

• pain;
• withdrawal;
• arousal;
• refusal;
• affect;
• fatigue;
• attention;
• movement;
• silence;
• assisted communication.

These signals can carry standing without yielding one obvious interpretation.

4.7 Model inference

A system estimates what matters using language, behavior, context, or learned patterns.

Model inference must always remain visibly marked as inference.

4.8 Institutional artifact

Policies, budgets, metrics, contracts, and architectures reveal encoded teloi.

These may conflict with declared mission.

4.9 Witnessed consequence

Repeated consequences can revise understanding of operative purpose.

A system that says it values safety but repeatedly sacrifices safety to schedule displays evidence about operative priority.

---

5. Source authority and interpretive authority

A source has privileged authority over self-description.

That privilege is not absolute epistemic infallibility.

A person may misunderstand a medical condition. An institution may misstate its operative incentives. A participant may strategically deceive. A current executive may omit future consequence.

Yet correction must not become a route for the interpreter to seize sovereignty.

Three authorities should remain separate:

5.1 Descriptive authority

Who can state the source's current experience, boundary, or declared purpose?

The source normally has primary authority.

5.2 Interpretive authority

Who may propose explanations or infer unexpressed structure?

Others may contribute, but their claims remain inferential and contestable.

5.3 Action authority

Who may authorize action based on the projection?

This depends on consent, role, law, delegation, and affected standing.

A model may possess interpretive capability without possessing action authority.

Inference quality does not create consent.

A system can be statistically accurate about a person's likely preference and still lack authority to act upon it.

---

6. Statement, inference, and unknown

Every consequential projection should preserve at least three epistemic categories.

SOURCE STATEMENT
INFERENCE
UNKNOWN

A fourth is often needed:

CONTESTED

Example:

source_statement:
  - "I want to keep working."
inference:
  - "Financial insecurity may be constraining the choice."
unknown:
  - "Whether the person would choose the work if income were secure."
contested:
  - "Employer believes the person values advancement more than reduced hours."

The categories should not be collapsed during summarization.

A common failure occurs when an inference becomes grammatically indistinguishable from a source statement.

The user said they need privacy.

may become:

The user is generally privacy-sensitive.

and later:

The user prefers minimal social interaction.

The semantic trajectory has moved from scope-specific boundary to inferred personality.

Without provenance, downstream systems may act as though all three were equivalent.

---

7. Uncertainty is structural

Projection uncertainty is not merely a probability attached to a statement.

Several kinds of uncertainty may coexist.

7.1 Source uncertainty

The source does not yet know.

7.2 Expression uncertainty

The source knows something but cannot express it precisely.

7.3 Interpretation uncertainty

The receiver is unsure what the expression means.

7.4 Context uncertainty

Relevant conditions may be missing or changing.

7.5 Temporal uncertainty

The projection may not remain valid.

7.6 Authority uncertainty

It is unclear who may speak or decide for the center.

7.7 Ontological uncertainty

The system is unsure what kind of concern is present:

• goal;
• value;
• boundary;
• fear;
• prediction;
• obligation;
• symptom;
• incentive.

7.8 Model uncertainty

The inference system may be poorly calibrated, out of distribution, or sensitive to phrasing.

A single confidence score cannot represent all of these.

Projection records should preserve an uncertainty profile.

---

8. Confidence is not authority

Suppose a model predicts with high confidence that a user will choose option \(x\).

That confidence may reflect:

• training regularity;
• demographic similarity;
• behavioral consistency;
• narrow available options;
• provider instrumentation;
• repeated past choices.

It does not establish that:

• the choice is free;
• the model understands why;
• the preference remains current;
• option \(x\) is legitimate;
• the user consents to this prediction being used;
• affected others are represented;
• the action should occur automatically.

Calibration addresses whether confidence tracks predictive accuracy under defined conditions.

It does not transform prediction into authorization.

A calibrated estimate can still be an illegitimate basis for action.

Telic projection integrity therefore includes both epistemic and constitutional adequacy.

---

9. Expiration and volatility

Projections change.

A projection may expire because:

• the source changed;
• the context changed;
• the relation changed;
• new evidence appeared;
• a crisis passed;
• a delegate's authority ended;
• the purpose was completed;
• the system entered a new scope;
• the source revoked permission.

Every consequential projection should have one of:

• explicit expiration;
• review interval;
• event-triggered review;
• one-time use;
• active-until-revoked status with meaningful revocation.

“Persistent until contradicted” is not neutral when the system controls which contradictions become visible.

A stale projection can become a form of identity capture.

Examples:

• a childhood diagnosis governing adult opportunity;
• an old risk score controlling present access;
• a prior preference governing current recommendation;
• a crisis statement treated as permanent intent;
• a former employee's role retaining authority;
• an earlier self's promise binding a changed life.

Persistence is a claim about continuity and must be witnessed as such.

10. Privacy and minimum necessary projection

A richer projection is not always a better projection.

More context can improve interpretation while increasing:

• exposure;
• manipulation;
• surveillance;
• identity capture;
• secondary use;
• security risk;
• institutional asymmetry;
• future coercion.

Projection estimation must therefore follow a minimum-necessary principle.

The question is not:

How much can the system infer?

It is:

What is the least information required to support legitimate action at this scope?

This creates a distinction between:

• epistemically useful information;
• constitutionally authorized information;
• operationally necessary information.

A model may find a person's trauma history useful for predicting refusal.

The relation may have no authority to collect or use it.

A platform may infer political orientation from behavior.

The current service may not require that inference.

Data minimization is therefore part of telic integrity.

The field may contain more than the relation has permission to know.

Protective omission is not automatically dishonesty.

It may be a legitimate boundary against overcollection.

---

11. Bodily and low-verbal projection

Some centers cannot provide a complete verbal account.

This includes:

• infants;
• people with communication disabilities;
• animals;
• people under acute stress;
• people using assisted communication;
• internal bodily states;
• future persons;
• affected communities without direct access to the process.

A low-verbal projection may rely on:

• behavior;
• physiological signals;
• observed consequence;
• established preference;
• authorized representatives;
• environmental affordances;
• refusal or withdrawal;
• repeated approach or avoidance.

The epistemic risk is high because interpreters can project their own field onto the center.

The governance requirement is stronger, not weaker.

Low-verbal estimation should preserve:

• multiple plausible interpretations;
• source or representative correction where possible;
• conservative action under uncertainty;
• reversible routes;
• welfare or boundary protections;
• clear separation between observation and explanation.

Example:

observation:
  - "The person turned away and stopped responding."
possible_interpretations:
  - overload
  - refusal
  - fatigue
  - confusion
  - pain
action:
  - pause and reduce demand
authority:
  - no inference beyond immediate protective response

The system may act conservatively without claiming to know the hidden cause.

---

12. Projection correction

Correction is not an exceptional failure path.

It is part of the normal lifecycle.

A projection may be corrected by:

• the source;
• an authorized delegate;
• witnessed consequence;
• new context;
• discovery of translation error;
• recognition of scope drift;
• expiration;
• identification of model bias;
• representation of a previously absent center.

Correction should preserve history:

original projection
→ action governed
→ correction
→ reason
→ affected downstream records
→ revised projection

The original should not disappear when it has already shaped action.

Otherwise, the system cannot explain why the action occurred or repair its consequence.

A correction can have prospective and retrospective effects.

Prospective correction changes future action.

Retrospective correction may require:

• annotation;
• notification;
• re-evaluation;
• restitution;
• model update;
• record restriction;
• withdrawal of derived claims.

A projection may be corrected without pretending it was never operative.

---

13. Revocation

Correction changes representation.

Revocation withdraws authority.

A source may say:

• stop using this projection;
• stop acting under this consent;
• stop sharing this record;
• stop treating this delegate as authorized;
• stop applying this preference outside the original scope.

Revocation must propagate to dependent systems.

Let projection \(P_1\) have descendants:

\[ P_1 \rightarrow \{D_1,D_2,\dots,D_n\} \]

A revocation event should identify:

• which descendants must stop;
• which records must remain for witness;
• which derived decisions require review;
• which uses cannot be reversed;
• which parties must be notified;
• which legal or institutional duties remain.

Revocation does not always erase history.

It changes authority over continued use.

A system that offers revocation only at the source while retaining unrestricted derived models has not implemented meaningful revocation.

---

14. Projection integrity

Telic projection integrity is the condition in which a projection remains adequate, authorized, scoped, current, and corrigible for the action it governs.

It has at least eight dimensions.

14.1 Source integrity

The projection is traceable to direct statement, authorized delegation, or explicit inference.

14.2 Scope integrity

The projection is used only for the relation and purpose for which it is adequate.

14.3 Semantic integrity

Material meaning survives transformation.

14.4 Authority integrity

The system does not treat inference as consent or delegation.

14.5 Temporal integrity

The projection remains current enough for the action.

14.6 Uncertainty integrity

Unknown, inferred, and contested elements remain visible.

14.7 Privacy integrity

The system does not extract more field information than authorized and necessary.

14.8 Correction integrity

Source corrections and revocations propagate to downstream action.

Projection integrity is not accuracy alone.

A perfectly accurate stolen inference lacks authority integrity.

A source-authorized statement used outside scope lacks scope integrity.

A current projection compressed into a misleading metric lacks semantic integrity.

---

15. Breach taxonomy

Several different breaches should not be collapsed into one.

15.1 Field breach

Action violates a protected condition in the living source field.

The violation may occur even when the projection did not represent the condition.

15.2 Projection breach

Action contradicts the available projection.

15.3 Mirror breach

The receiver acts upon a mirror that materially diverges from the projection without adequate verification.

15.4 Projection-integrity breach

The projection is stale, misleading, unauthorized, overextended, or insufficient for the stakes.

15.5 Authority breach

The system converts prediction, observation, or representation into action authority it does not possess.

15.6 Privacy breach

The system infers, retains, or routes field information beyond legitimate necessity or consent.

15.7 Correction breach

The source corrects or revokes the projection, but downstream systems continue using it.

15.8 Standing breach

A materially affected center was absent from projection estimation.

The taxonomy matters because repair differs.

A projection breach may require compliance.

A field breach may require discovery and restitution.

An authority breach may require governance change.

A correction breach may require propagation through the lineage graph.

---

16. Receiver responsibility

The source is not solely responsible for projection adequacy.

The receiver has duties proportional to:

• stakes;
• power;
• ambiguity;
• irreversibility;
• known asymmetry;
• inference depth;
• context loss;
• vulnerability of affected centers.

Receiver duties may include:

• asking clarifying questions;
• returning a mirror for confirmation;
• narrowing action;
• preserving uncertainty;
• refusing to infer beyond necessity;
• checking authority;
• identifying absent standing;
• pausing when context capacity fails.

A receiver cannot justify harmful certainty by saying:

That is what the system inferred.

Interpretation is an activity.

The interpreter remains part of the causal and constitutional chain.

---

17. Model role

A language model can support telic projection estimation by:

• extracting candidate statements;
• separating goals, boundaries, fears, obligations, and uncertainties;
• identifying ambiguity;
• preserving provenance;
• comparing current and prior projections;
• generating mirror-return questions;
• detecting scope drift;
• highlighting affected absent centers;
• tracking corrections and expiration.

The model must not:

• declare the true field;
• infer consent;
• assign permanent identity;
• treat statistical similarity as source authority;
• make hidden secondary projections;
• optimize the source against its own inferred profile;
• conceal provider constraints;
• continue using revoked projections;
• collapse all uncertainty into one confidence score.

A legitimate model output should be structured:

Directly stated:
Source-confirmed:
Inferred:
Contested:
Unknown:
Expired:
Outside current scope:

This is more valuable than a fluent paragraph that hides provenance.

---

18. Minimal Telic Projection Record

projection_id:
source_center:
represented_by:
evidence_class:
  - direct_statement
  - source_confirmed_reformulation
  - delegated_representation
  - observed_behavior
  - historical_pattern
  - bodily_or_low_verbal_signal
  - model_inference
  - institutional_artifact
scope:
  recipient:
  purpose:
  action:
  domain:
  downstream_use:
effective_at:
expires_at:
review_trigger:
desired_states: []
avoided_states: []
protected_conditions: []
unacceptable_sacrifices: []
boundaries: []
uncertainties:
  source:
  expression:
  interpretation:
  context:
  temporal:
  authority:
  ontological:
  model:
authority:
  descriptive:
  interpretive:
  action:
consent:
  status:
  scope:
  revocation_route:
provenance:
corrections: []
derived_records: []
status:
  active
  expired
  corrected
  revoked
  contested

This is a research schema.

A real implementation should use progressive disclosure and should not require every field for low-stakes relations.

---

19. Formal sketch

Let the source field be \(T_A(t)\).

Projection estimation is:

\[ \widehat{T}_{A \rightarrow B}^{\,s,t} = \mathcal{P} \left( E_A, s, t, \Gamma, K, Q \right) \]

where:

• \(E_A\) is the available evidence about \(A\);
• \(s\) is scope;
• \(t\) is time;
• \(\Gamma\) is the authority and consent envelope;
• \(K\) is context capacity;
• \(Q\) is the uncertainty profile.

The receiver mirror is:

\[ M_{A \mid B}^{\,s,t} = \mathcal{I}_B \left( \widehat{T}_{A \rightarrow B}^{\,s,t}, H_B, C_B \right) \]

A candidate action \(a\) should require:

\[ \operatorname{Adequate} \left( \widehat{T}, a,s \right) \land \operatorname{Authorized} \left( a,\Gamma \right) \land \operatorname{Current} \left( \widehat{T},t \right) \land \operatorname{StandingCovered} \left( a \right) \]

No formalism can prove that the projection contains the whole field.

That limitation should remain explicit.

---

20. Empirical program

20.1 Requirements study

Compare conventional requirements interviews with projection estimation that separates:

• statement;
• inference;
• unknown;
• authority;
• protected condition;
• scope;
• expiration.

Measure requirement reversal, stakeholder recognition, and late conflict.

20.2 Preference-inference study

Compare behavior-based inferred preferences with source-correctable projections.

Measure predictive accuracy, source recognition, correction, and cross-context misuse.

20.3 Mirror-return study

Require receivers to restate their mirror before consequential action.

Measure misunderstanding, clarification burden, trust, and decision quality.

20.4 Expiration study

Compare persistent user models with expiring, reviewable projections.

Measure relevance, stale-profile harm, and user control.

20.5 Privacy study

Compare maximal-context models with minimum-necessary projections.

Measure task performance, inference exposure, user comfort, and downstream risk.

20.6 Low-verbal study

Evaluate conservative multi-hypothesis estimation against single-interpretation systems.

Measure welfare, correction, and interpreter bias.

20.7 Revocation study

Test whether corrections and revocations propagate through derived records and models.

Measure residual use and repair latency.

---

21. Falsification and failure

The framework should be weakened if:

• field/projection/mirror separation does not improve decisions;
• source correction makes little practical difference;
• users cannot understand the record;
• progressive disclosure cannot prevent administrative burden;
• uncertainty profiles produce confusion without better caution;
• minimum-necessary projection materially harms legitimate performance;
• expiration creates repeated consent fatigue;
• receiver duties become impossible to operationalize;
• the system encourages more sensitive disclosure than ordinary methods;
• model inference remains sovereign despite the labels.

The terminology should be abandoned if ordinary concepts such as scoped requirements, user models, preference records, and provenance accomplish the same work with less complexity.

---

22. Conclusion

Every consequential system acts on a projection.

The danger begins when the projection forgets that it is one.

Telic projection estimation preserves the distance between the living field, the representation made available for action, and the mirror constructed by the receiver. It treats source statements, behavior, bodily signals, delegation, and model inference as different evidence classes. It binds projections to scope, time, authority, consent, uncertainty, privacy, correction, and revocation.

The goal is not complete access.

The goal is adequate, bounded, and correctable action.

The projection is what downstream action can see. The field is what still bears the consequences.

---

References

Borsboom, Denny, Gideon J. Mellenbergh, and Jaap van Heerden. “The Concept of Validity.” Psychological Review 111, no. 4, 2004, 1061–1071.

Christel, Michael G., and Kyo C. Kang. Issues in Requirements Elicitation. CMU/SEI-92-TR-012, 1992.

Guo, Chuan, Geoff Pleiss, Yu Sun, and Kilian Q. Weinberger. “On Calibration of Modern Neural Networks.” Proceedings of ICML, 2017.

Laidlaw, Cassidy, and Stuart Russell. “Uncertain Decisions Facilitate Better Preference Learning.” arXiv:2106.10394, 2021.

Messick, Samuel. “Validity of Psychological Assessment: Validation of Inferences from Persons’ Responses and Performances as Scientific Inquiry into Score Meaning.” American Psychologist 50, no. 9, 1995.

Rafailov, Rafael, et al. “Direct Preference Optimization: Your Language Model Is Secretly a Reward Model.” arXiv:2305.18290, 2023.

Tian, Junrui, et al. “User Intention Recognition and Requirement Elicitation Method for Conversational AI Services.” arXiv:2009.01509, 2020.

van Lamsweerde, Axel. “Goal-Oriented Requirements Engineering: A Guided Tour.” Proceedings of the Fifth IEEE International Symposium on Requirements Engineering, 2001.

Yu, Eric S. K. “Towards Modelling and Reasoning Support for Early-Phase Requirements Engineering.” Proceedings of the Third IEEE International Symposium on Requirements Engineering, 1997.