A claim is a statement that presents something as true, likely, valid, causal, evaluative or recommended.
“The train arrived at 8:05”, “this treatment reduces symptoms”, “the policy caused the decline”, “this source is reliable”, “this design is better” and “we should expand capacity” are all claims. They differ in type, scope, evidence burden, confidence, contestability and authority.
Quick answer: how should claims be categorised?
- Claim type: factual, causal, predictive, evaluative, normative, definitional, interpretive?
- Source: who or what asserts it?
- Scope: narrow, general, universal, population-specific?
- Evidence: what observations or sources support it?
- Confidence: certain, probable, possible, disputed?
- Testability: can it be checked or falsified?
- Time: when is the claim intended to hold?
- Jurisdiction: where does it apply?
- Provenance: where did the claim originate and how was it transformed?
- Contestability: what alternative claims compete with it?
This article extends How to Categorise Knowledge by treating claims as first-class objects that can be supported, challenged, revised and retired.
1. Separate claim from fact
A fact is a state of the world or a well-established proposition; a claim is an assertion made about the world. Claims can be true, false, uncertain or unresolved.
2. Separate claim from evidence
Evidence supports or challenges a claim. It should not be collapsed into the claim itself.
3. Separate claim from source
A source can contain many claims, and one claim can appear in many sources.
4. Factual claims describe states
They assert that something happened, exists, has a property or has a measurable value.
5. Causal claims connect change
They assert that one factor produces, contributes to or prevents another outcome.
6. Predictive claims concern future states
Forecasts should carry horizon, model, assumptions and uncertainty.
7. Definitional claims establish meaning
They state what a term means within a framework or community.
8. Interpretive claims explain significance
They connect evidence to meaning and often permit several defensible readings.
9. Evaluative claims compare quality
Better, worse, effective, fair and reliable require explicit criteria.
10. Normative claims state what should happen
They depend on values, obligations or principles and cannot be settled by empirical evidence alone.
11. Scope determines strength
“This worked in one trial” is narrower than “this always works”. Broad claims require broader evidence.
12. Universal claims carry the highest burden
One counterexample can defeat a genuinely universal proposition.
13. Population claims need population definition
Age, geography, context and sampling frame can determine whether a generalisation is valid.
14. Time-bounded claims should say when
A claim may be correct in 2026 and false after policy, technology or environment changes.
15. Jurisdiction-bounded claims should say where
Legal, educational and administrative claims often depend on local authority.
16. Source authority affects weight, not truth automatically
Experts and institutions can be wrong. Authority is evidence about reliability, not a substitute for the claim’s support.
17. Primary and secondary claims differ
A primary source may report direct observation while a secondary source synthesises or interprets other evidence.
18. Derived claims need lineage
Summaries, charts, models and AI outputs should point back to the underlying claims and evidence they transform.
19. Confidence belongs to the claim state
Confirmed, probable, possible, disputed and rejected describe epistemic status rather than claim type.
20. Confidence is not evidence quality
A claim can be held confidently on poor evidence. Track confidence and evidence quality separately.
21. Falsifiability matters for empirical claims
A useful empirical claim should identify observations that could count against it.
22. Some claims are not empirical
Logical, mathematical or normative claims require different forms of justification.
23. Contradictory claims should coexist during dispute
Do not overwrite one claim with another before the conflict is resolved. Preserve each source and evidence chain.
24. Competing hypotheses form a claim set
Diagnostic and scientific reasoning often improves when alternatives remain visible rather than forcing early closure.
25. Claim support can be direct or indirect
Direct observations and inferential chains should be distinguishable.
26. Corroboration needs independence
Ten sources copying one original assertion do not create ten independent confirmations.
27. Retractions change status, not history
A withdrawn claim should remain historically traceable with its current status and reason for withdrawal.
28. Superseded claims need successor links
A more precise later claim may replace an earlier broad claim without making the earlier record disappear.
29. Claim granularity matters
Large compound claims can hide which subpart is supported and which is disputed.
30. Split compound claims when evidence differs
If one clause can be independently tested, it may deserve its own claim ID.
31. Claims can be conditional
“If X, then Y” should preserve the condition rather than being indexed only by its conclusion.
32. Counterfactual claims need model assumptions
Statements about what would have happened under different conditions require an explicit causal model.
33. AI can extract claims
Models can identify propositions in text, but should preserve source, modality, negation and uncertainty.
34. AI can hallucinate claims
Generated assertions should not be treated as evidence merely because they are fluent.
35. Claim verification can be routed by type
Factual claims may need source lookup; mathematical claims may need proof; evaluative claims need criteria; predictions need later outcome comparison.
36. Claim taxonomies drift
New domains and media create new claim forms, including machine-generated and synthetic claims.
37. A practical claim record
- claim ID;
- claim text;
- claim type;
- source;
- scope;
- time;
- jurisdiction;
- supporting evidence;
- challenging evidence;
- confidence;
- status;
- competing claims;
- provenance;
- reviewer;
- version.
38. Claim categories should improve verification
If classification does not change what evidence or reasoning is needed, the category may not be useful.
39. Claims are movable pieces of knowledge
They can be proposed, tested, strengthened, challenged, narrowed, revised or retired while their provenance remains intact.
40. The deeper idea
Knowledge systems become more trustworthy when they distinguish what is asserted from what is observed and how strongly the two are connected.
To categorise a claim well is to preserve what is being asserted, by whom, over what scope, with what evidence and under what degree of uncertainty.
Final answer
Categorise claims by type, source, scope, evidence, confidence, testability, time, jurisdiction, provenance and contestability. Keep claims separate from facts and evidence, preserve competing claims during disagreement, and version claim status as evidence changes.