Editorial Principles

Human dignity is an evidence rule, not a soft tone preference. OHCHR reported verified civilian casualties in the first half of 2026, UNICEF reported UN-verified child casualties, WHO documented increased attacks on health care, and ICRC explains the legal obligation to allow humanitarian relief for civilians in need. Those records require language that treats people as civilians, children, patients, families and communities before it treats them as arguments.

Articles should therefore avoid spectacle, ethnic contempt, triumphal framing and casual images of suffering. A number can establish scale, but it cannot replace the legal and human status of the people counted. When the evidence is a verified minimum, the article must say so instead of converting it into a total story about all harm.

Source transparency means showing what a source proves and what it does not prove. OHCHR casualty reporting is a verified minimum, RDNA5 measures damage and needs through 31 December 2025, IOM displacement data has a collection period, and the Kiel tracker separates commitments, allocations and deliveries. These are not interchangeable facts.

Evidence-based articles must keep those boundaries visible in the sentence, not only in a source drawer. A reconstruction estimate should not become a final war cost. A verified casualty count should not become a total death toll. An aid allocation should not be described as delivered assistance unless the source uses that category.

Testimony must come from documented interviews, named reporting, verified records or clearly anonymised material with editorial notes. ICRC reporting on families of missing or captured people, a Scientific Reports study on trauma and resilience, and independent Ukrainian long-form journalism all show that war reaches private life in specific, documented ways. That is enough evidence; invented composite voices are not needed.

If a page uses a scene, quote, family detail or emotional description, it must be attached to a source or clearly labelled as analysis of a documented pattern. Fabrication is especially dangerous where people face repression, displacement or family separation, because a false scene can endanger trust in real testimony.

Corrections are part of evidence work because the evidence changes. OHCHR releases updated civilian casualty records, RDNA assessments are dated to a specific damage period, IOM displacement findings depend on collection windows, and the Kiel tracker updates aid categories and timing. A page that was accurate under one release can become incomplete under the next.

An evidence-based correction should identify the old wording, the new source or corrected interpretation, and whether the change affects the argument. Silent edits are acceptable only for spelling or formatting. Anything that changes a number, attribution, source category, uncertainty range or conclusion should be logged.

No collective blame follows from the evidence itself. Human Rights Watch reports repression against Russian critics, Amnesty describes centrally controlled school indoctrination, ICRC reports displacement and family tracing needs among Russian border-region civilians, and polling records personal exposure to grief, strain and economic hardship. These records do not erase state responsibility; they show why attribution must be precise.

Articles should attribute decisions to governments, armed forces, agencies, named officials or documented institutions where the source supports that attribution. They should not turn nationality, language, ethnicity or citizenship into guilt. Evidence about state policy is strongest when it does not punish civilians with lazy grammar.

Content labels must tell readers what kind of claim they are reading before they decide how much weight to give it. A verified OHCHR casualty update, the RDNA5 reconstruction-needs assessment, an S&P Global scenario pathway and a Ukrainska Pravda long-form story are all useful evidence, but they do different jobs. Treating them as the same genre would mislead readers even if every source is real.

Evidence-based articles should therefore label facts, analysis, scenarios, opinion, satire and testimony at the point of use. A verified minimum should not be written like a total. A scenario should not be written like a forecast. A reported story should not be inflated into a national sample. The label is part of the evidence trail, not a design extra.

Reviewer workflow should be tied to the risk of the article. Nuclear-safety text needs expert-source review because IAEA updates describe changing off-site power and cooling-water constraints without proving an accident is inevitable. Health-system articles need source review because WHO reports attacks and access limits without verifying every individual diagnosis or facility-level need. Humanitarian-law articles need legal-source review because ICRC explains obligations but does not prove compliance in every location.

A concrete workflow therefore assigns review tasks by evidence problem: source boundary, legal attribution, translation accuracy, conflict sensitivity, data category and update date. High-risk claims should not move from draft to publication until a reviewer has checked whether the source proves the sentence actually written.

Correction history should record when evidence changes or when the article used a source too broadly. OHCHR civilian-casualty releases update verified minimums, Kiel updates aid categories and timing, RDNA5 is bounded by damage and needs through 31 December 2025, and IOM displacement findings depend on a collection window. These sources are concrete examples of why a correct sentence can become stale or overbroad.

A correction entry should state the previous wording, the source that changed or clarified the claim, the corrected wording and whether confidence changed. It should not hide a changed number, category or date inside a silent rewrite. Readers need to know whether an update is a typo fix or a material change in evidence.

Satire separation is necessary because the underlying evidence concerns real deaths, missing families, repression and named military losses. OHCHR casualty records, ICRC reporting on families living with uncertainty, HRW reporting on repression and Mediazona named-death methodology all depend on reader trust in the boundary between documentation and invention.

Satire may criticise propaganda, official language or public hypocrisy, but it must not imitate verified reports, fake testimony, invented source badges or mock casualty ledgers. If a satirical item uses war evidence as context, it should label the satirical frame clearly and keep cited facts separate from invented elements.