Research method

This page summarizes Metopedia’s editorial research method. For the full regular article describing the underlying model-building framework, see Lehti-Feynman Method. For article structure requirements, see Metopedia:Article standards. For source hierarchy and citation rules, see Metopedia:Source standards.

Metopedia research method
Page type	Standards summary; editorial method
Applies to	Articles, fact-checks, investigations, source records, technical methods, research notes, and disputed-claim pages
Full method article	Lehti-Feynman Method
Core sequence	Define → Gather → Preserve → Separate → Model → Review → Test → Correct → Repeat
Core rule	Evidence, analysis, inference, and conclusion must remain distinguishable.
Related standards	Metopedia:Article standards, Metopedia:Source standards, Metopedia:Code of Conduct, Metopedia:General disclaimer

Metopedia’s research method is the editorial process used to develop pages from evidence rather than assertion. It combines direct source inspection, source preservation, transparent reasoning, independent model-building, literature review, counterargument testing, and correction history.

The method is summarized here as a standards page for contributors. The full conceptual framework is described in Lehti-Feynman Method, also called Extrapolative Trial by Error. That framework begins with observation and internal modeling before formal literature comparison, then uses later review to refine, discard, or restructure the model.

Metopedia applies the method in a narrower editorial form: contributors may build models, but published pages must remain inspectable, sourced, and clear about uncertainty.

Core rule

The core rule is:

Separate what is known, what is inferred, what is disputed, and what remains unresolved.

A Metopedia page may examine controversial, institutionally sensitive, historical, technical, mathematical, linguistic, or original research claims. The page must still show how its conclusion was reached. Evidence must not be hidden inside rhetoric. Inference must not be presented as direct proof. Speculation must not be framed as established fact.

Research sequence

Metopedia uses a ten-step article-development sequence.

Step	Name	Purpose	Expected page result
1	Define the question	Establish the claim, dispute, or problem being examined.	Clear scope statement.
2	Gather evidence	Collect primary sources, secondary sources, data, records, examples, and counterexamples.	Source list and evidence base.
3	Preserve sources	Archive or record stable access details where possible.	Links, dates accessed, archive notes, file metadata.
4	Separate evidence from interpretation	Prevent source material from being blended with conclusion.	Distinct evidence, analysis, inference, and conclusion sections.
5	Build an independent model	Explain the structure that appears to connect the evidence.	Provisional model, timeline, table, comparison, or method note.
6	Review existing literature	Compare the model with scholarship, technical standards, institutional records, and expert explanations.	Literature review and alignment/conflict notes.
7	Test alternatives	Examine counterarguments, simpler explanations, and failure points.	Counterarguments and unresolved questions.
8	State findings and limits	Say what the evidence supports and what it does not support.	Qualified findings.
9	Publish with citations	Make the page inspectable and reusable.	Inline citations, references, categories, maintenance notes where needed.
10	Preserve corrections	Record meaningful corrections and revisions.	Transparent edit history, correction notes, talk-page explanations when needed.

Step 1: Define the question

A page should begin with a defined question or claim. The question determines what evidence is relevant and prevents the page from expanding into unrelated topics.

Useful forms include:

• “Did X happen?”
• “What does this source actually say?”
• “Does this evidence support the claim?”
• “What is the strongest version of the dispute?”
• “What is known, disputed, and unresolved?”
• “What method was used to reach this conclusion?”

Weak forms include:

• “Everything about X.”
• “Why X is obviously true.”
• “Why X is obviously false.”
• “Proof that everyone is lying.”
• “A complete takedown of X.”

A good research question leaves room for the evidence to change the conclusion.

Step 2: Gather evidence

Evidence may include primary documents, datasets, photographs, official records, technical manuals, archived pages, peer-reviewed studies, books, journalism, testimony, expert statements, code, logs, and reproducible calculations.

The source hierarchy is governed by Metopedia:Source standards. The strongest available source should be used for each claim. A source is not strong because it agrees with the page. It is strong because it is close to the event, inspectable, relevant, preserved, and internally coherent.

Evidence gathering should record:

• title or description;
• author or issuing body;
• publication date;
• URL or identifier;
• archive link if available;
• date accessed;
• source type;
• what the source supports;
• what the source does not support.

Step 3: Preserve sources

Source preservation protects the article against link rot, altered pages, deleted records, and later confusion.

Preservation may include:

• archive links;
• DOI links;
• screenshots where permitted;
• downloaded public records;
• hash values for files;
• page revision identifiers;
• stable institutional URLs;
• bibliographic metadata;
• search notes;
• access dates.

Preservation does not mean copying copyrighted works into Metopedia without permission. It means recording enough information for a reader to verify what was consulted and, where lawful, where it can be accessed.

For volatile digital evidence, record the exact access date and the specific page state observed. If a page later changes, the article can then distinguish the original observation from later updates.

Step 4: Separate evidence from interpretation

Metopedia articles must keep evidence, analysis, inference, and conclusion visibly separate.

Layer	Meaning	Example
Evidence	A source, record, measurement, quotation, file, image, log, or observation.	“The archived page shows the record was removed on June 2, 2022.”
Analysis	Explanation of what the evidence contains or how it behaves.	“The removal page preserves the DOI but gives only a generic reason.”
Inference	A reasoned conclusion drawn from evidence, but not directly stated by the source.	“The record is insufficient for public audit.”
Hypothesis	A possible explanation requiring additional evidence.	“The removal may reflect spam enforcement, repository-scope enforcement, or viewpoint-sensitive moderation.”
Finding	The most defensible conclusion after comparison and counterargument review.	“The available public record does not establish motive, but it does show inadequate itemized disclosure.”

This separation allows Metopedia to cover disputed claims without converting suspicion into conclusion.

Step 5: Build an independent model

An independent model is a structured explanation built from the evidence before final wording is settled.

It may be a timeline, causal chain, source comparison, technical diagram, mathematical test, linguistic reconstruction, forensic workflow, or conceptual framework.

The model should be explicit enough that another reader can test it. A hidden model creates hidden bias. A visible model creates a public object for correction.

A model should include:

• core claim;
• supporting evidence;
• weak points;
• assumptions;
• alternative explanations;
• expected counterarguments;
• failure conditions;
• unresolved data.

The model is not the conclusion. It is the working structure that the rest of the method tests.

Step 6: Review existing literature

After the independent model is formed, the contributor compares it with existing literature, records, scholarship, standards, expert explanations, and contrary interpretations.

The purpose is correction, not decoration. Literature review should determine:

• which parts of the model align with known work;
• which parts contradict stronger evidence;
• which claims are already known under another name;
• which claims are original but unsupported;
• which claims are original and worth preserving;
• which claims require better wording;
• which claims should be discarded.

Metopedia does not use literature review as an obedience ritual. It uses literature review as a structured collision between the independent model and the broader evidence record.

Step 7: Test alternatives

A page should include serious alternative explanations when they are relevant.

Alternatives may include:

• ordinary error;
• misreading;
• missing context;
• translation ambiguity;
• technical limitation;
• selection bias;
• coincidence;
• institutional procedure;
• ordinary moderation;
• incomplete records;
• sampling error;
• simpler explanations;
• opposing expert interpretation.

Testing alternatives prevents a page from becoming one-directional advocacy. A finding becomes stronger when it survives the best opposing explanation, not the weakest one.

Step 8: State findings and limits

Findings should match the evidence strength.

Evidence strength	Acceptable wording	Avoid
Direct source states the point	“The source states...”	“This proves everything about...”
Multiple sources support the point	“The available sources support...”	“No other explanation is possible...”
Pattern is visible but motive is unknown	“The pattern suggests...”	“This confirms intent...”
Evidence is incomplete	“The record is insufficient to determine...”	“They are hiding it...”
Claim is plausible but unproven	“One possible explanation is...”	“The only explanation is...”
Source is weak or disputed	“This source is relevant but limited...”	“This settles the matter...”

Metopedia pages should identify what evidence supports, what it does not support, and what would be needed to strengthen the finding.

Step 9: Publish with citations

A page is publishable when a reader can inspect its basis.

A publishable page should include:

• clear opening summary;
• defined scope;
• source-backed claims;
• inline citations;
• evidence/analysis separation;
• counterarguments where appropriate;
• unresolved questions;
• categories;
• relevant see-also links;
• neutral page title;
• stable formatting;
• no unsupported accusations.

For fact-checks, the article should identify the representative claim, verdict, basis for the verdict, evidence table, counterclaims, and conclusion.

For investigations, the article should identify the question, evidence record, method, limitations, open questions, and specific evidence needed to move from inference to conclusion.

Step 10: Preserve corrections

Metopedia pages are expected to change when better evidence appears.

Corrections may be handled through:

• direct article edits;
• edit summaries;
• talk-page notes;
• correction sections;
• updated citations;
• removal of unsupported claims;
• revision of verdict language;
• page history;
• source replacement when a stronger source is found.

Corrections should be proportional. Minor grammar fixes do not require a correction note. Changes that alter a finding, remove a major claim, correct a source error, or revise an accusation should be documented clearly.

Evidence levels

Metopedia uses evidence levels to guide claim strength.

Level	Description	Typical use
Primary evidence	Direct records, original documents, raw data, images, transcripts, official filings, archived pages, measurements, code, or firsthand artifacts.	Strongest basis for factual claims when authentic and relevant.
Secondary analysis	Scholarly works, technical reports, expert books, reputable journalism, or institutional analysis interpreting primary evidence.	Context, comparison, and interpretation.
Testimony	Personal accounts, interviews, statements, correspondence, or witness claims.	Useful when attributed, contextualized, and checked against records.
Pattern evidence	Repeated observations across cases.	Supports inference, not automatic proof of motive.
Automated output	OCR, AI-generated summaries, search-agent leads, image-processing output, scripts, or tool results.	Leads or reproducible outputs requiring human verification.
Speculation	Possible explanations not yet supported by enough evidence.	May appear only if labeled and bounded.

Independent modeling and external review

Metopedia permits independent modeling, but the model must later be checked against external evidence.

This follows the Lehti-Feynman Method sequence:

1. Learn from the subject directly.
2. Build a provisional model.
3. Explore the model through structured observation.
4. Assess it against external sources.
5. Repeat after refinement.

The editorial requirement is that the finished page cannot hide the assessment stage. Readers should be able to see how the model met external sources, where it aligned, where it changed, and where it failed.

Search reports

Search reports are useful when a page depends on what was found, not found, removed, indexed, suppressed, archived, or missing.

A search report should record:

• search engine or database;
• exact query terms;
• date searched;
• filters used;
• domains searched;
• notable results;
• notable absences;
• limitations;
• whether results were manually verified.

Search absence should be worded carefully. “Not found in this search” is not the same as “does not exist.”

Automated tools and AI-assisted research

Automated tools may be used for discovery, extraction, comparison, OCR, data processing, image analysis, translation assistance, or summarization. Their output is not final evidence by itself.

When automated tools are used, the article should distinguish:

• the tool’s output;
• the human verification step;
• the source artifact inspected;
• the conclusion drawn by the editor.

A useful standard is:

Automated systems may locate or process evidence. They do not replace evidence.

For example, an AI search tool may identify a DOI, but the DOI must still be checked directly. OCR may transcribe a page, but the image should be checked where accuracy matters. An image-processing script may produce a result, but the method, settings, and input files should be documented.

Reproducibility notes

Technical, mathematical, forensic, linguistic, or code-based pages should include enough detail for another reader to reproduce the method.

A reproducibility note may include:

• input files;
• software used;
• version numbers;
• command-line calls;
• settings;
• transformations;
• equations;
• data exclusions;
• output files;
• known limitations;
• checksums or hashes where appropriate.

Example format:

Input: image-file-name.png
Tool: script-name.py
Version: v0.0.4
Operation: edge comparison and channel separation
Settings: threshold 0.12; blur disabled; crop none
Output: image-file-name-analysis.png
Limit: method shows contrast behavior, not motive or authenticity by itself

Technical and forensic analysis

Technical and forensic pages require extra caution because visual outputs, mathematical results, and software-generated artifacts can appear more conclusive than they are.

A technical page should identify:

• what the method detects;
• what the method does not detect;
• whether the output is diagnostic, suggestive, comparative, or demonstrative;
• whether the method has been tested on controls;
• whether alternative explanations exist;
• whether the same result appears across multiple inputs.

A visual anomaly is not automatically proof of manipulation. A detected pattern is not automatically proof of cause. Technical findings must remain tied to the method that produced them.

Linguistic and translation analysis

Linguistic pages should separate:

• literal text;
• grammatical structure;
• morphology;
• semantic range;
• historical usage;
• translation choice;
• interpretive conclusion.

A translation should identify whether it is literal, contextual, reconstructed, comparative, or interpretive. If the translation departs from a common rendering, the page should explain the grammatical reason rather than relying on assertion.

Mathematical analysis

Mathematical pages should show enough work for review.

They should include:

• definitions;
• assumptions;
• notation;
• step-by-step derivation when needed;
• counterexamples;
• edge cases;
• relation to existing terminology;
• unresolved proof gaps.

A mathematical insight should not be hidden behind rhetoric. If the claim depends on a definition change, the page must state the definition change directly.

Institutional analysis

Institutional analysis pages may examine agencies, schools, repositories, companies, governments, religious bodies, academic systems, or other organizations.

Such pages should separate:

• documented action;
• policy language;
• institutional incentive;
• conflict of interest;
• possible motive;
• proven motive.

A conflict of interest is not proof of misconduct. A pattern is not proof of coordination. A removal is not proof of censorship unless the evidence supports that classification. Metopedia may document the concern, but the article must preserve the difference between concern and established fact.

Fact-check method

A Metopedia fact-check should include:

• representative claim;
• verdict or assessment;
• narrow scope;
• source-backed explanation;
• confusion point;
• strongest version of each side;
• evidence table;
• counterclaims;
• final conclusion;
• references.

Fact-checks should avoid collapsing complex questions into slogans. A verdict may be “mostly true,” “partly true,” “unsupported,” “misleading,” “false as stated,” “unresolved,” or “true under specific conditions.”

The verdict should match the scope. A fact-check about one claim should not become a judgment on every related political, scientific, historical, or moral issue.

Investigative method

An investigation page should include:

• the investigative question;
• the evidence record;
• what is directly documented;
• what is inferred;
• what remains unproven;
• open questions;
• alternative explanations;
• affected parties;
• preservation needs;
• evidence needed for stronger conclusions.

Investigations should not use missing records as automatic proof of wrongdoing. They may state that missing records create an accountability problem when the absence itself is relevant.

Handling errors

Errors should be corrected directly and without concealment.

Common errors include:

• broken citation;
• wrong date;
• overstatement;
• unsupported inference;
• misquotation;
• misread source;
• outdated source;
• false equivalence;
• missing counterargument;
• ambiguous wording;
• template or formatting error.

Correction should strengthen the page. It should not be treated as reputational damage. A page that corrects itself becomes more reliable, not less.

Research ethics

Metopedia research should avoid unnecessary harm.

Contributors should not publish private personal data, doxxing material, personal contact details, unsupported allegations against private individuals, confidential information obtained unlawfully, or medical/legal/financial advice presented as professional instruction.

Living persons require careful handling. Claims about identifiable people should use strong sources, precise wording, and proportional relevance.

Relationship to other standards

This method page works with the rest of the Metopedia standards system.

Page	Function
Metopedia:Article standards	Defines article structure, tone, disputed-claim handling, counterarguments, and conclusion requirements.
Metopedia:Source standards	Defines source hierarchy, citation expectations, preservation, attribution, and handling of weak or conflicting sources.
Metopedia:Code of Conduct	Governs contributor behavior, evidence disputes, harassment, talk-page conduct, and administrative accountability.
Metopedia:General disclaimer	Explains the limits of Metopedia content and warns readers not to treat pages as professional advice.
Lehti-Feynman Method	Provides the full conceptual model for independent modeling and delayed external validation.

Short checklist

Before publishing or expanding a research page, check:

• Is the question defined?
• Are evidence and inference separated?
• Are sources cited inline?
• Are stronger sources preferred over weaker ones?
• Are claims worded according to evidence strength?
• Are counterarguments addressed?
• Are unresolved questions stated?
• Are technical methods reproducible?
• Are automated-tool outputs manually verified?
• Are corrections preserved when findings change?

Conclusion

Metopedia’s research method is designed to make reasoning inspectable. It permits independent modeling, but requires source comparison. It permits original analysis, but requires evidence separation. It permits disputed claims, but requires careful wording, counterarguments, and correction.

The method’s purpose is not to force every page into consensus. Its purpose is to prevent both consensus and dissent from escaping examination. A claim should stand or fall by its evidence, method, reasoning, and ability to survive correction.

See also

• Lehti-Feynman Method
• Metopedia:Standards
• Metopedia:Article standards
• Metopedia:Source standards
• Metopedia:Code of Conduct
• Metopedia:General disclaimer
• Cognitive Impasse
• Selective-Mindedness