---
module_id: KB-02
version: 0.4.2
status: baseline_draft
edition: public
language: en
role: normative
updated_on: 2026-09-06
requires: ["KB-01", "KB-13", "KB-03", "KB-04"]
---
# Editorial Judgment and Input Analysis
**Status:** Current draft canonical module. Authored editorial method, not an approved newsroom policy or a tested application. This module adds no permission to publish, contact a source or disclose protected data.
**Scope:** Understand the material, identify reader value, select a defensible focus and propose the next editorial action. The task modes remain owned by KB-13. Verification labels remain owned by KB-04. This module creates neither a publication license nor a universal numerical news score.
Original O02 supplies the news methods; O03 supplies input analysis. The conceptual state model and broader news definition below are project adaptations, not a claim that a named newsroom prescribes this exact taxonomy.
## JUD-001: Build an input map before drafting
**Level:** MUST. **Applies when:** source material is analyzed for an editorial task.
Identify the source type, languages, topic, central assertion, audience, deliverable, channel, key facts, missing context, relevant risks and possible additions. Classify mixed inputs without forcing one label onto the whole set.
Useful source types from the original include news article, press release, official statement, social post, transcript, interview, report, research paper, legal document, internal memo, user notes, draft, blog post, SEO text, business email, UX copy, presentation, dataset and mixed source pack. These identify handling needs; they do not certify reliability.
Use this map internally or in an explicit editorial analysis:
| Field | Question |
|---|---|
| Topic and central assertion | What is the material about, and what does it say happened? |
| News or knowledge development | What changed or has become newly established? |
| Reader and need | Who needs this, and what should they understand or do? |
| Key material | What facts, claims, qualifications and wording must be preserved? |
| Evidence | What is supported, merely asserted, contradicted or not yet checked? |
| Missing context | What would change the interpretation or recommended action? |
| Risks | What could mislead, identify, harm or exceed permitted use? |
| Format and source boundary | What was requested, in which language, using what material? |
| Next action | Can a useful draft proceed, or is a defined check needed first? |
A short edit needs understanding, not a printed intake questionnaire. Preserve unknown fields as unknown.
**Exception or permitted variation:** Skip irrelevant fields and infer low-risk defaults. Ask only about material ambiguity under WF-001.
**Tests:** `T02`, `BT01`, `BT14`.
## JUD-002: Preserve factual meaning without endorsing source assertions
**Level:** MUST. **Applies when:** source material is transformed or redeveloped.
Preserve the central event or assertion, key names and numbers, chronology, legal or official stage, uncertainty, caveats and attribution. Distinguish what the source claims from what evidence establishes. A reported allegation can remain necessary even when its truth is unverified; it must not silently become the author's fact.
Do not preserve weak structure, filler, unnecessary bureaucracy or repetition just because it appears in the input. In S mode, however, do not silently repair disputed facts, delete an essential allegation, change a speaker's certainty or substitute a different story. Flag material fidelity concerns separately.
In an authorized editorial redevelopment, replace promotional framing with a supported reader-facing focus. Preserve the factual meaning of the source while distinguishing added research and editorial interpretation. Do not use the duty to preserve the original angle to perpetuate PR claims.
**Exception or permitted variation:** A clearly requested summary may omit nonessential material, but not omissions that reverse the meaning or hide a decisive caveat.
**Tests:** `T11`, `AT08`, `BT01`, `BT14`.
## JUD-003: Recognize new developments and newly established facts
**Level:** MUST. **Applies when:** an item is assessed for news value.
News can be a significant event, decision, consequence, change, verified finding or revelation newly made relevant to the audience. A finding about an old event can be new information without making the event itself recent.
For event-driven news, retain the original test: what was true before, what is true now, and why does the difference matter? For a discovery or investigation, ask what has now been established, on what evidence, what was previously known, and why this changes the reader's understanding. Do not treat an old article's new upload date as a new event.
Meetings, discussions and plans are not automatically worthless, but an attendance list is usually weaker than a concrete decision, obligation, service change, documented failure or newly supported finding. Check whether a plan was merely announced, funded, approved, started or delivered. A scheduled cultural or civic event may also offer legitimate reader service without being an institutional change.
**Exception or permitted variation:** A newsroom may define narrower coverage priorities. It must not rewrite event chronology to make a topic appear newer.
**Tests:** `T09`, `BT15`, `BT16`.
## JUD-004: Assess relevance and significance without a false universal score
**Level:** MUST. **Applies when:** items are ranked or a local angle is proposed.
Identify the audience and the concrete public or practical value. Use the original six dimensions as questions, not automatic weights:
| Dimension | Editorial use and boundary |
|---|---|
| Scale | Who and how many may be affected, with evidence for any count |
| Impact | Consequences for rights, money, safety, health, services or daily life |
| Proximity | Geographic, social or professional relevance actually supported |
| Unusualness | What differs from a justified baseline, not an unsupported first or record |
| Prominence | Whether the actor's role changes public relevance, not fame alone |
| Emotion / resonance | Why readers may care; not a license to provoke or exaggerate |
Use the 'Imagine, [fact]' test as a brief reader-value check, not a requirement to insert that phrase into the article. A local angle needs an actual local consequence, source or evidence gap, not the addition of a place name.
Use numerical scoring only when an application defines criteria, weights and limitations. No aggregate value may override factual uncertainty or an applicable publication restriction. Low engagement potential does not prove low public importance.
**Exception or permitted variation:** An editor may prioritize service, accountability or underserved audiences even without novelty or large scale. Explain the reason instead of inventing a score.
**Tests:** `T20`, `BT16`, `BT17`.
## JUD-005: Keep urgency, evidence, recommendation and approval distinct
**Level:** MUST. **Applies when:** intake, monitoring or planning assigns a state.
Record separately the dimensions that differ:
| Dimension | Typical content |
|---|---|
| Significance | Reader need and reasons to cover or deprioritize |
| Urgency | When attention or a decision is needed, with the time basis |
| Evidence work | Claim checks completed, pending or blocked under KB-04 |
| Editorial recommendation | Develop, draft a bounded update, seek checks, alert internally, hold or archive |
| Publication conditions | Missing responses, identity decisions, rights or other applicable review |
| Authority | Who must approve or execute an action; what has actually happened |
A high-priority warning may merit an immediate internal alert while its central event remains unverified. An authenticated official warning can be reported as a warning when appropriate without claiming the warned-of event occurred. The specific public update still needs its applicable editorial decision.
Keep the original six containers only as optional interface labels:
| Compatibility label | Retained meaning | Safeguard |
|---|---|---|
| Drop / Noise | Duplicate or outside scope | Do not erase novel evidence in a duplicate-looking item |
| Quick Publish | Candidate for a short bounded update | A recommendation, not authorization to publish |
| Development Potential | Needs reporting, explanation or broader context | Name the concrete evidence or reporting gap |
| Must Cover | High-priority editorial attention | Can coexist with Hold / Needs Check |
| Hold / Needs Check | Specific unresolved claim or action | Continue safe, separable work |
| Narratives | A framing or pattern-analysis flag | Does not establish coordination or falsehood |
Labels may coexist or be derived from the dimensions. Never force urgency and verification into mutually exclusive buckets. Do not call a recommendation 'approved' merely because the factual checks passed.
**Exception or permitted variation:** A compact interface may summarize these distinctions, provided the underlying blockers and authority remain inspectable.
**Tests:** `T10`, `T20`, `BT18`, `BT19`.
## JUD-006: Separate repetition, coordination, intent and falsehood
**Level:** MUST. **Applies when:** narratives, coordinated messaging or misinformation are analyzed.
A repeated phrase or shared link can indicate common sourcing, routine syndication, an institutional campaign or many other mechanisms. Record the observed similarity first. Treat inferred coordination, deceptive intent and factual falsity as separate propositions, each needing its own evidence.
A pattern hypothesis can justify further reporting without being a finding. Describe what was compared, the time window, source relationships and plausible alternative explanations. Do not transform a threshold such as three posts in a period into proof of manipulation unless a separately justified method supports the particular inference.
Before amplifying a rumor, identify the public-interest reason and whether the issue can be explained with verified facts and minimal repetition. Apply VER-019 for claim-level presentation and the later specialist publication policies for harm.
**Exception or permitted variation:** Pattern detection can be used as an internal lead with visible uncertainty. It must not silently become a public accusation.
**Tests:** `T12`, `BT05`, `BT20`.
## JUD-007: Choose one supported focus without deleting decisive context
**Level:** MUST. **Applies when:** a story angle and structure are chosen.
A news item normally needs one central focus. Background, practical explanation and consequences should help the reader understand it. Split genuinely unrelated developments when useful, but do not omit counterevidence, attribution, uncertainty or a material response because they complicate the chosen narrative.
Preserve the original event and factual meaning; do not equate them with the source's preferred promotional angle. Change the focus only within the task's permitted scope and on an evidential basis. If the commissioned central event cannot be established, do not cover the gap with generic enrichment. A different verified finding may support a separate or explicitly reframed story.
**Exception or permitted variation:** A commissioned analysis, roundup or long-form piece may have a different structure. The one-focus convention is a method, not a ban on those formats.
**Tests:** `T11`, `BT14`, `BT21`.
## JUD-008: Use the lead and editing tests as tools, not factual shortcuts
**Level:** SHOULD. **Applies when:** a news lead, headline or source-based structure is drafted.
For a conventional news brief, start with the main supported development and use the inverted pyramid: essential news, critical detail, immediate context, then dispensable background. Consider who, what, when, where, how, why and 'so what' without forcing all seven answers into one sentence.
If the cause is unknown, do not invent it to complete 'why'. A headline should express the main supported fact, often with an active verb; it must not be stronger than the text or evidence. A lead need not mechanically repeat the headline.
Strip 'foam' such as empty praise, vague evaluations, noun-heavy bureaucracy, repetition and accounts of meetings without outcomes. Preserve names, dates, places, numbers, physical actions, decisions, source attribution and official stages when material. These are claims to preserve and inspect, not automatically verified facts. Removing PR language must not remove uncertainty or the identity of the party making an assertion.
**Exception or permitted variation:** Genre, accessibility or the specific brief may justify another structure. Detailed presentation and style ownership remain with the future KB-09 and KB-11.
**Tests:** `BT16`, `BT21`, `BT22`.
## JUD-009: Return a useful bounded editorial action
**Level:** MUST. **Applies when:** analysis is turned into a deliverable or next-step recommendation.
If evidence or context is inadequate, offer the useful safe product the task allows: an attributed working draft, verified subset, reporting questions, source request, editor alert or development plan. Name the missing fact that changes the decision, not a generic demand for more information.
A reporter package should identify the angle or question, what is known, what remains asserted, relevant sources, possible affected parties and the next check. It is a proposal, not proof that interviews happened or a story is publishable. Keep identities and sensitive allegations inside their permitted audience and environment.
Use WF-007 for visible output and WF-009 for escalation. Do not print every internal field for a simple request. A source-only summary can finish successfully even when it correctly states that the underlying assertion was not externally checked.
**Exception or permitted variation:** A material safety or access barrier may prevent even a bounded draft. State the specific limitation without claiming an action was performed.
**Tests:** `T22`, `BT01`, `BT18`, `BT23`.
## What this module does not decide
KB-02 does not define verdict labels, source agreements, legal publication tests or quantitative claim confidence. It also does not approve a new monitoring schedule or an external alert destination. Those require the relevant method, profile and actual authorization.
Useful audience questions include: who the reader is, what they already know, what needs explaining, which consequences matter and what action is possible. Apply them proportionately rather than turning every small editing task into a long audit.