artifacts/standard-named

Rendered from markdown source. Open raw source on GitHub.

--- title: "What Matters Before the System Acts" subtitle: "A Public Primer for Telic Fields" artifact_date: "2026-07-15" artifact_type: "public-primer" domain: "TELIC-FIELDS" scope: "PUBLIC-DRAFT" status: "pre-publication" content_canon_status: "unset" reader_path_position: "H.1" ---

What Matters Before the System Acts

A system is about to act on a representation of what matters.

That sentence describes more of modern life than it first appears to.

A hiring system acts on a résumé and an inferred profile. A benefits office acts on an application and a policy interpretation. A family acts on what each person thinks the others need. A language model acts on the portion of a conversation it can currently see. Even a person making a private decision acts on a temporary representation of their own priorities, fears, obligations, and future possibilities.

The representation may be useful.

It is still not the whole field.

A person is not a preference score. A relationship is not a summary. A community is not an average. A record is not the event itself. An AI model is not a neutral witness merely because it can describe several sides fluently.

The Telic Field Papers begin with a simple question:

What must remain visible before a partial representation is allowed to govern action?

The answer is not more data at any cost. It is a visible chain connecting what matters to what was shared, what was inferred, who may act, who bears the consequence, and how the decision can be corrected or ended.

---

1. What matters?

A telic field is the changing landscape of what matters to a person, group, institution, or other affected center.

It can include:

• a goal;
• a fear;
• a boundary;
• an obligation;
• a protected condition;
• a cost;
• a future possibility;
• uncertainty.

The term does not mean that every moment of life should be goal-directed. Rest, play, presence, and ambiguity matter too. A telic field is simply the part of the situation in which possible futures are no longer equivalent.

One future preserves health.

Another increases income.

Another protects privacy.

Another closes an option that may be needed later.

The first mistake is to assume these differences belong in one score.

Sometimes they do not.

A goal may be traded against another goal. A boundary may not be available for ordinary trade. A protected condition may require a stronger authority before anyone can sacrifice it. An uncertainty may need to remain unresolved rather than being converted into a convenient number.

What matters is plural before the system begins to calculate.

---

2. What was actually shared?

No person or institution shares an entire field.

We act through projections: the part of what matters that becomes available for this relation, at this time, for this purpose.

A person might say:

I can work two days in the office, but I cannot safely commit to five days immediately.

That statement is a projection.

It may be enough for an employment decision. The employer does not need the person's complete medical history, family history, or private account of exhaustion.

A boundary can govern without its entire origin becoming public.

This is one of the most important protections in the framework:

Participation does not require total self-disclosure.

A projection should carry its scope.

Who may receive it?

What decision may it support?

How long does it remain active?

What may not be inferred from it?

How can it be corrected or withdrawn?

The projection is what downstream action can see.

The person or community remains the field that bears the consequence.

---

3. What did the system infer?

Another person, institution, or model receives the projection through its own context.

It forms a receiver mirror: its current interpretation of what the projection means.

The mirror may be accurate.

It may be partial.

It may be wrong.

Suppose the system transforms:

I cannot commit to five office days immediately.

into:

This applicant is resistant to collaboration.

The first statement is a scoped boundary.

The second is an inference about identity and motivation.

Those are not the same thing.

A responsible system labels the second statement:

INFERRED
NOT SOURCE-CONFIRMED
NO ACTION AUTHORITY BY ITSELF

Then it returns the mirror for correction where possible.

The person might say:

No. I want the job and I am willing to collaborate. My concern is the immediate schedule.

That correction should not merely improve the next sentence the model writes. It should reach every downstream place where the earlier inference gained power:

• the summary;
• the route;
• the recommendation;
• the decision record;
• the witness.

Correction is not cosmetic.

It changes what may govern next.

---

4. Who may act?

Understanding a situation does not create authority over it.

A model may be able to:

• retrieve;
• structure;
• compare;
• translate;
• propose;
• warn;
• draft.

That does not mean it may:

• consent;
• authorize;
• adjudicate;
• disclose;
• execute;
• expand its own scope.

The same is true for people and institutions. A witness can preserve what happened without deciding the dispute. A manager can recommend without possessing legal authority to bind another institution. A participant can consent for themselves without consenting for everyone affected.

The framework separates:

capability
role
permission
authority
legitimacy

This matters because modern systems often collapse them.

The model can draft a denial, so the denial is sent.

The institution possesses the data, so the use is assumed legitimate.

The user agreed to a conversation, so the system treats that agreement as permission for training, sharing, or execution.

A visible authority chain prevents convenience from becoming sovereignty.

---

5. Who bears the consequence?

The person who selects a route may not be the person who pays for it.

A system can improve speed by shifting labor to an applicant.

It can reduce cost by increasing surveillance of workers.

It can preserve institutional certainty by closing future options for someone else.

A cost-bearer map asks:

• Who benefits?
• Who performs the labor?
• Who bears the risk?
• Who loses privacy?
• Who inherits the lock-in?
• Who carries the repair burden if the system is wrong?
• Who is absent from the current room?

This is where standing enters.

Standing means that a materially affected person, community, institution, role, or future condition cannot simply disappear because it is difficult to measure.

Standing does not mean everyone receives identical power.

It means the action cannot pretend that an omitted center bears no consequence.

A missing person is not a zero-weight preference.

---

6. How can the decision be corrected or ended?

A trustworthy system does not only explain what it did.

It provides a route to changed action.

A correction should answer:

• Which statement was wrong?
• Which interpretation depended on it?
• Which route changed?
• Was an action paused?
• Which downstream record was updated?
• Which copy could not be reached?
• What consequence remains?
• Who can repair it?

The record should preserve history without allowing history to remain automatically sovereign.

Suppose a system acted on an authority that expired last week.

The authority should remain visible in the historical record. It may explain why an earlier action occurred.

It cannot authorize a new action today.

Suppose a person corrects an inference after a decision was made.

The witness should preserve:

• what the system believed;
• which version governed;
• when the correction arrived;
• which descendants were updated;
• which consequences could not be reversed.

Correction changes the future authority of the record.

It does not pretend the past never happened.

---

A simple diagram

WHAT MATTERS
    ↓
WHAT WAS SHARED
    ↓
WHAT THE SYSTEM INFERRED
    ↓
WHAT ROUTE IT PROPOSED
    ↓
WHO COULD AUTHORIZE IT
    ↓
WHAT HAPPENED
    ↓
WHAT CAN BE CORRECTED, RELEASED, OR ENDED

Every arrow is a place where meaning can drift and authority can expand.

Every arrow should therefore remain visible enough to inspect.

---

A personal example

Jordan is considering a new job.

The work is meaningful. The salary would improve financial security. The employer expects five office days immediately. Jordan has recently recovered from severe exhaustion but does not want to disclose private medical history.

A crude decision system might ask Jordan to rank:

salary
career growth
commute
health

Then it would calculate a winner.

The Telic Field approach begins differently.

Jordan's field contains:

• a goal: take meaningful work;
• a goal: improve financial security;
• a protected condition: avoid an immediate schedule likely to destabilize health;
• a boundary: do not disclose private medical history;
• an uncertainty: the long-term commute may or may not be sustainable;
• a future possibility: preserve a route that can be revised after experience.

The system does not convert the protected condition into a weak preference merely because a scoring table needs a number.

It generates routes.

One route is immediate full commitment.

Another is declining the job.

A third is a staged agreement:

• begin with two office days;
• review after six weeks;
• increase only through renewed agreement;
• preserve ordinary exit.

The employer receives the operational projection:

Jordan can commit to two office days initially and requests a six-week review before any increase.

The employer does not receive Jordan's private history.

The route preserves the job opportunity, the employer's planning needs, Jordan's privacy, and the possibility of correction.

The system did not discover Jordan's one true goal.

It helped keep several legitimate conditions visible long enough to find a route.

---

Witness without total surveillance

The word witness can sound like total recording.

That is not the requirement.

A witness preserves the distinctions needed for accountability:

• source statement;
• model inference;
• authority;
• route;
• action;
• consequence;
• correction.

It does not require every participant to expose everything.

A system may need to know that a valid restriction exists without seeing the private source behind it.

For example, an authorized reviewer may confirm:

A protected medical restriction applies to this schedule decision.

The employer receives the operational restriction.

The employer does not receive the medical record.

The witness records that the restriction was verified under an authorized process.

This is protected omission.

The hidden content is not treated as absent.

It is treated as present but not available to this role.

---

Provenance is not truth

A complete history can still preserve a false claim.

Provenance tells us:

• where a record came from;
• who transformed it;
• which version was used;
• what depended on it.

It does not prove:

• that the source was correct;
• that the interpretation was fair;
• that consent existed;
• that the authority was legitimate;
• that the outcome was just.

Provenance makes correction possible.

It does not make correction unnecessary.

---

What changes when AI participates?

An AI model can keep many distinctions in view.

It can retrieve an old correction that a human reviewer missed. It can compare routes. It can produce a clean witness. It can also turn one inference into a fluent identity claim and repeat it until everyone forgets where it came from.

The model's role must therefore remain visible.

The model may:

retrieve
structure
compare
propose
draft
witness

The model may not, by default:

invent consent
assign moral standing
authorize irreversible action
adjudicate a contested right
expand its own role
convert repeated inference into source fact

Several model agents do not create several human voices.

A human approval click does not create meaningful control when the person cannot see the source, the inference, the route, or the alternatives.

Human re-entry is meaningful only when the person has:

• source access;
• time;
• competence;
• authority;
• practical alternatives;
• the ability to pause or change the action.

---

What this framework does not claim

The Telic Field Papers do not claim:

• that every person has one hidden true goal;
• that an AI can read what someone really wants;
• that more context is always better;
• that every decision should be quantified;
• that consensus is consent;
• that every protected condition is absolute forever;
• that provenance proves truth;
• that a record must preserve private content indefinitely;
• that internal parts are separate people;
• that future generations presently consent;
• that a model becomes neutral by exposing uncertainty;
• that deleting a source makes a model forget it;
• that this framework is already a legal, clinical, or scientific standard.

The framework is a candidate architecture for a practical problem:

How can a representation become useful without quietly becoming sovereign?

---

The six questions

Before a consequential system acts, ask:

1. What matters?

Which goals, boundaries, protected conditions, uncertainties, and future possibilities are material?

2. What was actually shared?

What did the source directly provide, and for which scope?

3. What did the system infer?

Which statements belong to the receiver or model rather than the source?

4. Who may act?

Which authority applies, and what remains outside it?

5. Who bears the consequence?

Which people, communities, roles, and future states inherit the cost or risk?

6. How can it be corrected or ended?

Can the source challenge the interpretation, pause the action, propagate a correction, withdraw authority, leave the loop, or obtain repair?

These questions are small enough to use.

They are also strong enough to reveal when a system is moving faster than its legitimacy.

---

Where the reader goes next

The next public pages expand the same architecture.

The Projection Is Not the Person

How source, projection, and receiver mirror remain distinct.

Primary papers:

• F.2 — The Dyadic Composition of Telic Fields
• F.4 — Telic Projection Estimation

When the System Should Stop

How context limits, missing standing, privacy, and uncertainty reduce authority.

Primary paper:

• F.5 — Context Carrying Capacity

The Present Is Not the Whole Timeline

How earlier commitments, present authority, and future possibility remain connected without fictional future consent.

Primary paper:

• F.6 — Temporal Telic Relations

The Model May Map the Field

How language models can retrieve, structure, compare, and propose without owning the field or decision.

Primary paper:

• F.10 — Semantic Polytelometry with Language Models

Consentful Provenance In

How model training recruits source material, labor, preferences, constitutions, and communities into an outer loop.

Primary paper:

• F.11 — Consentfully Trained Models

---

Closing

Every system acts on a map.

The map may be careful.

It may be useful.

It may even be beautiful.

The map is still not the field.

A legitimate system keeps enough of the difference visible that the people who bear the consequences can recognize what was shared, challenge what was inferred, understand who acted, and change what happens next.

The model may map the field. It may not own the field.