The Reference Framework

Design patterns, source-criticism rules, and fallacy-anticipation methodology for source-linked reference pages.

This page documents the framework — design patterns, source-criticism rules, and fallacy-anticipation methodology — on which every Factual Foundations topic rests. It is not about any specific topic. For worked examples, see the Volhynia report and the Portal proposal.

Purpose and Scope

The Factual Foundations reference library has three kinds of pages:

  1. Topic pages — full factual record on a specific contested issue (e.g. the Volhynia report). Topic pages apply this framework to a specific question.
  2. Proposal pages — explain why the reference library exists, what its standards are, and how it could scale (e.g. the Portal proposal). Proposal pages argue for the format.
  3. Methodology pages — this page. It is framework-only, with no topic content. It documents the design patterns, the source-criticism rules, and the fallacy handbook that every topic page should apply.

If you are trying to understand a specific issue, read a topic page. If you want to know why the reference library works this way, read a proposal page. This page is for readers who want to evaluate the methodology itself before reading any specific topic.

Design Patterns for Honest Reporting

Five design patterns. Every Factual Foundations topic page applies all five.

🔍

1. Full-Context Sourcing

Saying "they killed 100,000 people" hides the severity. Include the specific methods, not just the aggregate count. A reader who only sees a number has no way to judge the moral weight of what happened.

📐

2. Read-Across Methodology

Admit every source has editorial constraints. Show readers how to triangulate across sources rather than picking one. Cross-edition comparison (PL/EN/UK Wikipedia on the same topic) is the strongest example.

⚖️

3. Source-Bias Disclosure

For every outlet referenced, disclose known biases with citations to the bias studies themselves. Manhattan Institute 2024, AER 2012, Wikipedia's own systemic bias page are all citable sources.

⚠️

4. Uncomfortable-Information Rule

A page that avoids uncomfortable specifics is sanitized, not neutral. Present the facts that explain the reaction, even when they are graphic. Use content warnings, not omission.

🚫

5. No False Equivalence

Do not balance documentation of documented violence with the offender's framing of why the violence was necessary. Document the violence. State the framing. Let the reader weigh both — but do not artificially "balance" by quoting the framing alongside the documentation as if they had equal weight.

Source Criticism Framework

Every outlet referenced in a Factual Foundations topic has its known editorial bias documented. The framework has three layers.

Layer 1 — Genuine bias research (use the actual studies)

When a topic is contested, cite real bias research rather than guesswork. Examples:

Layer 2 — Cross-edition comparison (when the topic exists in multiple languages)

For any topic with multiple Wikipedia editions (most history topics, current events, people), read all three major editions. The differences in coverage are themselves evidence of editorial intent. On a topic like the Volhynia massacres:

Always start with the most detailed language edition. Treat the least-detailed as the official-narrative sample and weight accordingly.

Layer 3 — Known-limitations disclosure (per outlet)

Every outlet referenced on a Factual Foundations topic has its known limitations documented. The standard table:

Source typeBest forMissing
Wire services (AP / Reuters)Dependable "what happened" with named attributionAlmost no historical or causal context
Public broadcasters (BBC)Balanced framing, open methodologyMay omit uncomfortable specifics for tonal reasons
Independent journalism (Notes from Poland)Detailed chronology, embedded primary sourcesCountry-centric framing; less balanced toward opponents
Government institutes (IPN / UINM)Forensic primary-source documentationGovernment-aligned framing; treat as the institutional voice, not neutral
Wikipedia (any language)Sourced synthesis across multiple language editionsSystemic left-bias (EN), language-nationalist framing (PL/UK/RU/UA)

The Fallacy Handbook: Anticipating Dismissals

The 21 fallacies below are the canonical abstract patterns of the dismissal moves FF counters. The example quotes in the table are taken from FF reports - currently the Volhynia page is the canonical worked example, but each fallacy can be applied to any topic with a similar pattern.

How to use this handbook on a new topic:

  1. Identify which fallacies the topic will face. The 21 are common - most contested topics trigger 5-10 of them. Skim the names; the ones that match your predicted dismissal moves are the ones to pre-empt.
  2. Replace the example quote with one from your topic. The pattern is fixed (e.g. "What about [unrelated action by the other side]?"); the example is per-topic. The abstract pattern stays; the actors change.
  3. Keep the counter abstract enough to apply. The methodology counters are written to apply to any topic with the same pattern. They name the rebuttal type, not the topic-specific facts.
  4. Update the methodology TABLE only when the canonical pattern changes. If a new FF topic surfaces a fallacy move that doesn't fit the 21, add it to the methodology table and reference it from the topic page.
# Name Pattern Counter
1 Genetic Fallacy "Generated by [AI / hostile actor], so we cannot trust it." Attack the specific claim or accept the cited source. Who wrote it doesn't change what's true.
2 Poisoning the Well "That's just [foreign propaganda / opposition talking points]." Who benefits from a claim is independent of whether it is true. Present documented evidence.
3 Guilt by Association (Ad Hominem) "That's what [party X] says, so it is partisan." Test a claim by its sources, not by who repeats it. Truth isn't a popularity contest.
4 Whataboutism "What about [unrelated action by the other side]?" Expands the frame to other events; the rebuttal narrows it back. Even if the other action happened, it doesn't erase what's documented here.
5 Sanitisation by Reduction "What about [positive trait of the actor]? They were [heroic role]." Honouring a complex figure means reckoning with their whole record. One good trait doesn't cancel out documented crimes.
6 Loaded Question "Anyone who criticizes [actor] must be [discredited group], right?" A claim stands or falls on its own, whatever the speaker's politics. Independent sources agreeing is evidence — not a conspiracy.
7 Circumstantial Ad Hominem "Why now? / bad timing / [motivated by political calendar]." When a fact is cited is not the fact itself. The historical record doesn't change based on who brings it up or when.
8 Tu quoque "The critic has [tainted past / hypocrisy], so they have no standing to speak." The speaker's standing and the claim's truth are different things. Even with a tainted past, the facts remain what they are.
9 False Equivalence "That was [context: wartime, both sides, etc.]." Equivalence requires proof, not assertion. Spontaneous chaos is not the same as planned operations with documented methods.
10 Reverse Ad Populum "Polls show that [X]% disagree, so the claim is just opinion." A poll measures sentiment, not history. Sentiment can be wrong. The cited evidence stands on its own.
11 Overuse of Context "You have to see it from [actor]'s perspective! / Context!" Context helps you understand; it doesn't retroactively justify. You can't erase events by appealing to context — especially not to soften the impact of what happened.
12 Appeal to Consequences "Saying this helps [bad actor] / undermines [righteous cause]." Political inconvenience doesn't make something false. Pointing out consequences is not the same as refuting the fact.
13 Model-Bias Objection "[AI Model X] is biased — don't trust what AI says." Genetic fallacy at one remove. The AI merely confirmed what the sources already say. Engage the sources, not the tool.
14 Motte and Bailey "First [extreme claim]; after challenge, retreat to 'we only meant [moderate claim]'." State the original claim clearly. The arguer cannot put forward an extreme claim and, after criticism, retreat to an easier-to-defend one.
15 Straw Man "So you're saying [exaggerated / oversimplified version of the claim]." Restate the original claim exactly. A straw man is a simplified or extreme distortion — knocking it down doesn't touch the real argument.
16 Policing Tone "Even if it's true, the way you/they are saying it is too aggressive." Tone is not evidence. Politeness doesn't make something true or false. Engage with the content, not the delivery.
17 Bulverism "Of course you/they are saying [claim]; you are [position], so you are biased." Guessing at motives is not refuting. Investigating motives with evidence is valid; assuming them as a substitute is pre-emptive dismissal.
18 No True Scotsman "True [group members] wouldn't do [action]; it's only a marginal element." Reject the purity test. If people claiming that identity committed the action, appealing to the identity doesn't erase it.
19 Ad Hominem (direct) "You are [trait / motive / qualification], so your argument is suspect." Attack the claim, not the person. Character and motives don't alter the evidence. Every claim stands or falls on its own.
20 Argumentum ad Logicam (fallacy fallacy) "Your argument uses a logical fallacy, so the conclusion is false." A flawed argument doesn't make the conclusion false. Judge the conclusion on its own merits and the independent evidence.
21 volhynia F15 Argument from Negative Results "Searched [place], found no [thing] — so [thing] doesn't exist." One empty search doesn't erase documented history. The event is attested by independent sources — 80-year-old evidence is hard to locate. Absence of proof is not proof of absence.

On the Use of an LLM to Compile Reports

Every Factual Foundations report is compiled by an LLM. Which model — see the version history at the end of this section. The predictable dismissal: "It was written by AI, therefore it's unreliable / hallucinated / propaganda." That is a genetic fallacy (judging a claim by its origin rather than its content) and facts dismantle it quickly.

A more recent variant — common on X — is: "Grok is biased." Used as a counter when someone else has used Grok (or another AI) to fact-check a Factual Foundations claim, and the AI confirms the claim. Dismissing the AI's confirmation because "the AI is biased" is genetic fallacy at one remove. The model returned the same conclusion the report did — that is not bias, that is independent confirmation of the external, human-produced sources cited. The right next move is to engage those sources, not to dismiss the tool. See fallacy #13 in the handbook above for the full counter.

Fact: Every factual claim in every Factual Foundations report has an external, human-produced source. The LLM did not generate any of the following from its training data:
  • The specific methods of UPA violence — sourced to Wikipedia's article (citing Motyka, Snyder, Davies, Polish underground reports) and IPN publications
  • The eyewitness account of Tadeusz Piotrowski — directly cited, with book and page reference
  • The 2017–2024 exhumation ban — sourced to Notes from Poland and Ukrainska Pravda reporting
  • The 22 June 2026 Kancelaria Prezydenta statement — sourced to Kyiv Independent, RBC-Ukraine, and Ukrainska Pravda
  • The media-bias research — sourced to Manhattan Institute and AER papers
The LLM's contribution: format, language, and prose assembly. The claims themselves: external, human, cited.

Model Version History

The specific LLM used to compile each Factual Foundations report is documented here. Reports cite the model in their meta-bar at the top of the page; the version history records changes over time.

ModelPeriod usedNotes
DeepSeek V4 Flash Jun–Jul 2026 (initial pages) First model used to compile the FF portal and three reference pages. Model name was shown explicitly in early LLM-disclosure sections.
MiniMax M3 From Jul 2026 (current) Switched to this model in July 2026. All subsequent updates, corrections, and re-deploys use this model. The fact-checking methodology is identical: every claim is cited, the LLM only formats and retrieves.
DeepSeek V4 Flash Jul 2026 Polish translation of all FF pages (this translation).
Why this section exists. The genetic-fallacy dismissal ("it's AI, therefore it's unreliable") doesn't depend on which model was used. Whether the report was written by DeepSeek V4 Flash, MiniMax M3, or any future model, the validity of the cited facts is unchanged. This section is provided so a reader can see the full record, not because the model choice is itself a quality signal. A human researcher using the same tools would produce the same content.

When to Read This vs. A Topic Page

Use this decision tree:

📋

If you want to understand a SPECIFIC issue

Read a topic page. The Volhynia report is the canonical example — full factual record on a specific contested issue.

🚀

If you want to know WHY the format exists

Read the Portal proposal. It argues for self-updating, source-linked, bias-disclosed, fallacy-audited reference pages as an alternative to mainline news.

🔧

If you want to evaluate the methodology

You're reading it. This page exists so readers can assess the framework before applying it to any specific topic. It documents the patterns, the source-criticism rules, and the fallacy handbook.

References

Studies and sources cited in this methodology page:

Why no links? Sources are listed by name only, not as hyperlinks. Three reasons: (1) Wikipedia entries are editorially controlled by the same community that shapes the bias we are documenting - linking to a Wikipedia article amplifies its SEO weight and effectively endorses it as a source; (2) several sources we name here (Ukrainian Wikipedia, Wire services with state bias) are cited specifically because they fail some standard - linking to them is a contradiction; (3) the cited primary sources (Motyka's book, IPN archive, BBC, Notes from Poland) are accessible by name search in any browser, so the link is unnecessary for navigation. If you find a source you cannot locate by name, that is a finding worth noting in the report itself.
Manhattan Institute (2024)
Research
"Is Wikipedia Politically Biased?" 28,000+ article computational study.
American Economic Review (2012)
Peer-reviewed
First empirical measurement of Wikipedia political slant.
Wikipedia — Systemic Bias
Self-documentation
Wikipedia's own documentation of its systemic biases, including the NPOV policy and the dispute over its application.
Media Bias / Fact Check
Bias research
Independent media-bias rating service. Used for outlet-level bias scoring on the Volhynia page's source criticism section.
IPN (Instytut Pamięci Narodowej)
Government
Polish government institute for the documentation of 20th-century history. Primary source for the Volhynia excavations, archival evidence, and official Polish government statements.
UINM (Ukrainian Institute of National Memory)
Government
Ukrainian counterpart to IPN. Provides the institutional Ukrainian perspective on the same events; used to surface the Ukrainian national-narrative framing.
Grzegorz Motyka
Academic
Polish historian, leading scholar on the Volhynia massacre. Author of "Od rzezi wołyńskiej do Akcji Wisła" (From the Volhynia massacre to Operation Vistula). Primary source for UPA methods, the liquidation order, and the Polish response.
Tadeusz Piotrowski
Eyewitness / Academic
Polish-American sociologist, eyewitness to the Polish-Ukrainian conflict. Author of "Genocide and Rescue in Wołyń" — direct eyewitness account of UPA methods.