Basic Logic Fact-Check Method

Standard

This method prevents a captured transcript, a suggestive screenshot, or a list of links from being mistaken for a completed fact-check. It is designed for contentious claims about race, genetics, crime, intelligence, history, and policy, where variable definitions and causal levels matter as much as whether a quoted number appears in a source.

Project navigation

• Research hub — project status, conclusions, and queue
• Video inventory — every public upload and review status
• Cross-video claim map — recurring inference chains and reused evidence
• Source and link catalog — links requiring verification and their primary-source replacements

Status vocabulary

Only complete means fact-checked. A transcript-ready note is a research queue item, not an evidentiary conclusion.

Workflow

1. Preserve the source record

• Record exact title, URL, video ID, upload date, duration, description, chapters, and review date.
• Save the available automatic-caption transcript without silently rewriting substantive wording.
• Label automatic captions because they can mishear names, citations, numbers, and punctuation.
• Preserve screenshots of every consequential table, paper figure, quotation, map, and chart.
• Record the timestamp and whether the evidence appears onscreen, in narration, in the description, or only in comments.

2. Build the claim ledger

Break narration into claims small enough to evaluate independently:

• descriptive statistic;
• measurement claim;
• source attribution;
• causal claim;
• historical claim;
• analogy;
• generalization;
• policy or moral conclusion.

Do not let one true premise lend credibility to an untested conclusion bundled into the same sentence.

3. Recover the primary source

For each cited or visually identifiable source:

• locate the original paper, dataset, official report, statute, judgment, or archival document;
• identify the exact table, figure, page, population, year, sample size, variable, and model;
• read the methods, limitations, and authors’ own interpretation;
• distinguish the published version from a blog summary, screenshot, preprint, or later review;
• record corrections, retractions, replications, and major methodological critiques.

For scientific questions, later reviews can contextualize a study but do not replace checking the cited study itself.

4. Name the variables and level of inference

Before evaluating a conclusion, write down what was actually measured.

5. Audit the causal chain

Ask at every arrow:

1. Is the proposed cause measured?
2. Is the proposed outcome measured?
3. Is temporal order established?
4. Are plausible confounders addressed?
5. Does the design distinguish inherited effects from correlated environments or social treatment?
6. Does the evidence concern individuals, families, local populations, national aggregates, or historical institutions?
7. Is the inference transported to a different population, time, or outcome?
8. What finding would falsify the claim?

Use a visible chain when a video moves across multiple levels:

measured result → biological interpretation → psychological trait → group outcome → policy

Mark each arrow separately. Never give the whole chain the credibility of its strongest premise.

6. Audit visuals

• Compare the video’s labels, legend, scale, crop, colors, and annotations with the original.
• Check how much variance a PCA or dimensionality-reduction plot displays.
• Identify omitted categories, withheld test samples, altered labels, and absent uncertainty.
• For a map, recover the underlying observations, imputations, sample years, coverage, and aggregation rules.
• For a meme or screenshot, trace it back until the official or primary source and transformation are known.
• Preserve both the video’s frame and the source figure where possible.

7. Investigate provenance without overclaiming

Rate links by evidence:

Use wording such as “possible source,” “probable lineage,” “shared citation ecosystem,” or “direct use verified.” Do not infer a creator’s identity from topic, nationality clues, timestamps, or writing similarity alone.

8. Seek contrary and updating evidence

• Read serious critiques and serious defenses.
• Prefer research designs capable of distinguishing the competing causal stories.
• Search for replications, larger or more representative samples, family-based designs, and revised estimates.
• Record contrary evidence even when it weakens the working hypothesis.
• Treat a conclusion as having survived the current audit, not as permanently proven.

Verdict key

• Supported: the claim matches the relevant source and is appropriately scoped.
• Mostly supported: the main claim is sound but needs a limited qualification.
• Partly supported: a real result is present, but important scope or interpretation is wrong.
• Misleading: omitted context, labels, denominators, sampling, or measurement materially changes the impression.
• Unsupported inference: the source does not test or entail the conclusion.
• False under the ordinary meaning: the claim contradicts the relevant evidence or confuses distinct quantities.
• Unverifiable from available evidence: neither confirmation nor falsification is presently justified.

Verdicts attach to individual claims, not to a person’s character or an entire ideological identity.

Topic-specific checks

Genetics and intelligence

• Separate within-group heritability from causes of between-group means.
• Check population stratification, indirect parental genetic effects, assortative mating, and portability across ancestry groups.
• Distinguish gene annotation from a measured phenotype.
• Distinguish race as a social classification from genetic ancestry and the exact sampled populations.
• Do not use a clear monogenic adaptation as stand-in evidence for an unmeasured highly polygenic behavioral difference.

Crime statistics

• Repair malformed links and recover the official table.
• Distinguish arrests, offenders, offenses, convictions, prisoners, victims, and victimization surveys.
• Use the correct denominator and compare like years, jurisdictions, offenses, age structures, and reporting coverage.
• Check repeat arrests and whether the table contains unique persons.
• Separate descriptive disparity from the causal claim used to explain it.
• Treat a racialized headline such as “entirely driven by” as a claim to test, not a summary of the dataset.

History, law, and civil rights

• Start with the primary statute, judgment, executive action, speech, or archival record.
• Separate legal holding, political motive, enforcement, implementation, and later consequences.
• Do not infer the whole movement from one figure, faction, quotation, or outcome.
• Use contemporaneous context and note when terminology changed.

Development and civilization

• Test institutional, geographic, epidemiological, demographic, colonial, trade, conflict, and human-capital explanations.
• Avoid treating modern continental labels as stable historical actors.
• Distinguish descriptive correlation from historical causation.
• Require a mechanism and comparative cases before attributing complex institutions to ancestry.

Standard video-note structure

## Research status and source record
## Executive verdict
## Inference chain
## Timestamped claim-by-claim fact-check
## Study or source audits
## Onscreen evidence
## Methodological errors
## Source provenance
## Cross-video connections
## What evidence would change the verdict?
## Open questions
## Source register
## Full transcript

Obsidian title convention

The filename and frontmatter title already identify the note. Do not begin a note with an H1 that merely repeats that title. Start the visible body with an abstract, navigation callout, or the first substantive ## section. Use an H1 only when it adds a genuinely different document-level label.

Reference routing by topic

Completion checklist

• Metadata and full transcript preserved
• Caption uncertainties marked
• Consequential screenshots preserved with timestamps
• Every central claim appears in the ledger
• Original sources and exact locations recovered
• Variables, samples, denominators, and comparison groups named
• Descriptive, causal, and policy claims separated
• Counterevidence and serious alternative explanations reviewed
• Provenance confidence calibrated
• Cross-video source and argument links added
• No opening H1 merely duplicates the filename or frontmatter title
• Open questions and falsifiers recorded
• Note status changed to complete only after all preceding checks

Related notes

• Basic Logic Youtube Channel Research
• Basic Logic Youtube Channel Inventory
• Basic Logic Cross-Video Claim Map
• Basic Logic Source and Link Catalog