Claims and Propositions

The smallest unit of knowledge that can stand on its own and be judged as true or false — the atom from which all larger arguments, lessons, and knowledge systems are built.

What is it?

When you read a textbook chapter, listen to a lecture, or scan a Wikipedia article, you are absorbing dozens of individual statements bundled together. Some are facts (“Water boils at 100 degrees Celsius at sea level”). Some are interpretations (“The Industrial Revolution improved average living standards”). Some are definitions (“A prime number is a natural number greater than 1 that has no positive divisors other than 1 and itself”). Each of these is a claim — a single assertion that can be evaluated as true, false, or uncertain.

In philosophy, this idea has a formal name: a proposition. A proposition is the meaning behind a declarative sentence — the thing you are asserting to be the case.¹ Two different sentences can express the same proposition (“It is raining” and “Rain is falling” assert the same thing). The proposition is the atom of knowledge, independent of how it is worded.

In education, the same idea appears under a different name: atomisation. Kristopher Boulton’s framework identifies three types of atomisation — routine (breaking down step-by-step processes), factual (isolating individual facts), and conceptual (decomposing abstract ideas into their component concepts).² Each atom is, at its core, a single claim: one thing that can be taught, tested, and remembered independently. A teacher who has not atomised their material is asking students to learn compound structures without understanding the parts.

In computer science and knowledge engineering, the claim maps directly to a semantic triple: subject-predicate-object. “Paris is the capital of France” becomes (Paris, is-capital-of, France). This is the fundamental unit of a knowledge graph — a single relationship between two entities.¹ Every knowledge graph, every ontology, every structured database is built from these atoms.

Questions deserve attention here too. A well-formed question — “Is spaced repetition more effective than massed practice?” — is structurally identical to a claim. It is a proposition whose truth value is unknown. Treating questions as first-class atoms, stored and linked with the same rigour as answers, turns them from passive gaps into active tools for directing future learning and research.³

In plain terms

A claim is to knowledge what a brick is to a building. You can hold a single brick, examine it, test its quality, and decide where it belongs. You cannot do any of those things with a wall — you have to work brick by brick. Claims are the bricks of understanding.

At a glance

From messy knowledge to atomic claims (click to expand)
graph LR
    A[Messy Knowledge] -->|Decompose| B[Individual Claims]
    B -->|Evaluate| C{True / False / Unknown}
    C -->|True| D[Verified Knowledge]
    C -->|False| E[Rejected]
    C -->|Unknown| F[Open Question]
    D -->|Structure| G[Knowledge Graph]
    F -->|Investigate| B
Key: Unstructured knowledge is decomposed into individual claims. Each claim is evaluated for its truth value. True claims become verified knowledge that can be structured into graphs and systems. Unknown truth values become questions — atoms that drive further inquiry.

How does it work?

1. The claim as the unit of evaluation

A claim is the smallest statement you can meaningfully judge as true or false. “The French Revolution began in 1789” is a claim. “The French Revolution” is not — it is a topic, not an assertion. “The French Revolution was complex and changed many things” is technically a claim, but it is too vague to evaluate meaningfully. Good claims are specific and testable.¹

The discipline of identifying claims forces precision. When you encounter a paragraph that says “AI is transforming healthcare in remarkable ways,” you should ask: what exactly is being claimed? That AI is used in healthcare? That it improves outcomes? That adoption is increasing? Each of those is a separate claim, and each requires different evidence.

Think of it like...

A scientist designing an experiment. You cannot test “Is this medicine good?” You have to isolate a specific, testable claim: “This medicine reduces blood pressure by at least 10 mmHg within 4 weeks in adults with hypertension.” The sharper the claim, the more useful the test.

2. Separating claims from evidence

One of the most important disciplines in knowledge work is keeping claims and their supporting evidence as separate objects, not fused together in a single sentence.⁴ When you write “Spaced repetition improves retention by 200% (Cepeda et al., 2006),” you have welded a claim to a citation in a way that obscures important questions: How strong is this evidence? Does it replicate? Does it apply in all contexts?

Separating the two reveals hidden problems. A bold claim supported by a single small study looks very different from a modest claim supported by a meta-analysis. When claim and evidence are stored separately, you can see the proportionality between what is asserted and what supports it.⁴

This separation also enables reuse. A single piece of evidence may support multiple claims. A single claim may be supported by multiple, independent pieces of evidence. Fusing them into one note destroys these relationships.

Example: claim-evidence separation (click to expand)

Fused (harder to evaluate): “Retrieval practice is more effective than re-reading for long-term retention (Roediger & Karpicke, 2006).”

Separated (easier to evaluate):

Component Content
Claim Retrieval practice is more effective than re-reading for long-term retention
Evidence Roediger & Karpicke (2006): two experiments with undergraduates, 1-week delay, free-recall test. Retrieval group recalled 61% vs. re-reading group 40%.
Strength Two controlled experiments, consistent results, widely replicated since

Now you can assess the claim independently: is the evidence strong enough? Does it generalise beyond undergraduates? Does it hold for recognition tasks, not just free recall?

Component	Content
Claim	Retrieval practice is more effective than re-reading for long-term retention
Evidence	Roediger & Karpicke (2006): two experiments with undergraduates, 1-week delay, free-recall test. Retrieval group recalled 61% vs. re-reading group 40%.
Strength	Two controlled experiments, consistent results, widely replicated since

3. Atomisation in pedagogy

Boulton’s framework gives teachers a systematic method for breaking curriculum content into teachable atoms.² The three types each address a different kind of knowledge:

Routine atomisation breaks a multi-step procedure into individual steps. For example: solving a quadratic equation involves (1) recognising the form, (2) identifying coefficients, (3) applying the formula, (4) simplifying. Each step is a separate atom that can be taught and practised in isolation.
Factual atomisation unpacks the concepts embedded in a factual statement before presenting the fact itself. You cannot teach “Two is the smallest prime number” unless students already understand “prime number” and “smallest.”⁵
Conceptual atomisation decomposes abstract ideas into their component concepts. Understanding “democracy” requires understanding “voting,” “representation,” “majority rule,” and “rights” — each an atom in its own right.

The insight common to all three: you cannot teach a compound structure until you have taught its atoms. Attempting to do so causes the cognitive overload that makes students feel a subject is “just too hard.”²

Think of it like...

Building with LEGO. You cannot hand a child a completed castle and expect them to understand it. You hand them individual bricks, show how they connect, then build up section by section. Each brick is an atom. The castle is the compound knowledge.

4. Claims as semantic triples

In knowledge engineering, the claim takes a precise computational form: the semantic triple. Subject — predicate — object. “Paris is-capital-of France.” “Water boils-at 100C.” “Spaced repetition improves retention.”¹

This mapping is not accidental. The reason semantic triples are the foundation of knowledge graphs is that they are the computational equivalent of claims — the smallest meaningful unit of structured knowledge. Every entry in Wikidata, every fact in Google’s Knowledge Graph, every relationship in a medical ontology is a triple, which is a claim, which is a proposition.

Concept to explore

See semantic-triples for how subject-predicate-object structures power knowledge graphs, linked data, and the semantic web.

5. Questions as atomic units

A question like “Does interleaving improve transfer learning?” has the same structure as a claim — it asserts a relationship between concepts — but with an unknown truth value. Treating questions as first-class atoms means storing them with the same structure, links, and metadata as claims.³

This has practical consequences. A question stored as an atom can be linked to the claims it relates to, the evidence that partially addresses it, and the other questions it depends on. It becomes a tool for organising future learning, not just a prompt for recall.³

Key distinction

A claim asserts something and invites evaluation. A question identifies something unknown and invites investigation. Both are atomic. Both deserve the same structural treatment in a knowledge system.

Why do we use it?

Key reasons

1. Precision. Decomposing knowledge into claims forces you to state exactly what you believe, what you know, and what you are uncertain about. Vague understanding hides behind compound statements; atomic claims expose it.¹

2. Testability. A single claim can be verified, falsified, or qualified. A paragraph cannot. The discipline of isolating claims is the discipline of making knowledge amenable to scrutiny.⁴

3. Teachability. Students learn more effectively when material is broken into atoms they can master individually before combining them into larger structures. Atomisation reduces cognitive overload and makes sequencing transparent.²

4. Machine readability. Knowledge systems — from databases to knowledge graphs to RAG pipelines — operate on atomic units. A claim maps cleanly to a semantic triple, a database row, or a retrieval chunk. Compound prose does not.¹

When do we use it?

When building a knowledge base and deciding what the fundamental units should be
When teaching or explaining a complex topic and determining how to sequence the material
When evaluating an argument and separating what is claimed from what is supported
When designing a curriculum and identifying the prerequisite atoms for each lesson
When structuring content for RAG and choosing how to chunk documents so that each chunk carries a complete, evaluable assertion
When writing clearly and wanting to ensure every paragraph makes identifiable, defensible claims

Rule of thumb

If you cannot state a single, testable claim that your paragraph is making, the paragraph probably needs to be decomposed further — or you have not yet clarified what you actually think.

How can I think about it?

The legal testimony analogy

In a courtroom, a witness is asked to testify about specific facts — not to deliver a speech. “Did you see the defendant at the location on March 5th?” is a request for a single, evaluable claim. The lawyer breaks the narrative into atomic assertions, each of which can be cross-examined, corroborated, or challenged independently.

The witness’s full story = unstructured knowledge

Each specific question = isolating one claim

The answer (yes/no/I don’t know) = establishing the truth value

Cross-examination = testing the claim against evidence

The jury’s verdict = synthesising many evaluated claims into a conclusion

A legal system built on “tell us everything you know” would be chaos. A knowledge system built on compound, unevaluated statements is similarly unreliable.

The chemistry analogy

Chemists do not study “stuff.” They decompose substances into molecules, molecules into atoms, and atoms into subatomic particles. Each level of decomposition reveals different properties and enables different operations. You cannot understand water’s behaviour from its appearance; you need to know it is H2O — two hydrogen atoms bonded to one oxygen atom.

A substance = a body of knowledge (a chapter, a lecture, a domain)

A molecule = a compound claim (a paragraph making several related assertions)

An atom = a single claim or proposition

Chemical bonds = the logical relationships between claims (implies, contradicts, supports)

The periodic table = a taxonomy of claim types (facts, definitions, causal claims, evaluative claims)

Just as chemistry became a science only when it learned to work at the atomic level, knowledge work becomes rigorous only when it learns to identify and handle individual claims.

Concepts to explore next

Concept	What it covers	Status
semantic-triples	Subject-predicate-object structures as the computational form of claims	stub
knowledge-granularity	The design question of what level of decomposition preserves meaning	stub
constructivism	The learning theory that knowledge is built by assembling pieces, not received whole	stub
structured-data-vs-prose	The trade-off between machine-parseable structure and human-readable narrative	complete

Some cards don't exist yet

A broken link is a placeholder for future learning, not an error.

Check your understanding

Test yourself (click to expand)

Explain what a claim is and why it matters that claims can be evaluated as true or false. Use an everyday example.

Name Boulton’s three types of atomisation and give one example of each from a subject you know well.

Distinguish between a claim and its evidence. Why is it important to store them separately in a knowledge system?

Interpret this scenario: a teacher presents a lesson on climate change as a single 30-minute narrative without breaking it into individual claims. What problems might students experience, and how would atomisation help?

Connect claims to semantic triples. How does the structure of a claim (subject-predicate-object) enable knowledge graphs to reason about relationships?

Where this concept fits

Position in the knowledge graph
graph TD
    DIKW[DIKW Hierarchy] --> CP[Claims and Propositions]
    DIKW --> KE[Knowledge Engineering]
    CP --> ST[Semantic Triples]
    CP --> KG[Knowledge Granularity]
    style CP fill:#4a9ede,color:#fff
Related concepts:

semantic-triples — the computational form of a claim, mapping subject-predicate-object to nodes and edges in a knowledge graph

knowledge-granularity — the design problem of choosing the right level of decomposition, where claims represent the finest meaningful unit

constructivism — the learning theory that frames knowledge as built from smaller pieces, with claims as the building blocks

structured-data-vs-prose — claims sit at the boundary between structured data (triples, rows) and prose (paragraphs, narratives)

Yiuno

Explorer

Claims and Propositions

Claims and Propositions

What is it?

At a glance

How does it work?

1. The claim as the unit of evaluation

2. Separating claims from evidence

3. Atomisation in pedagogy

4. Claims as semantic triples

5. Questions as atomic units

Why do we use it?

When do we use it?

How can I think about it?

Concepts to explore next

Check your understanding

Where this concept fits

Sources

Further reading

Graph View

Table of Contents

Backlinks

Yiuno

Explorer

Claims and Propositions

Claims and Propositions

What is it?

At a glance

How does it work?

1. The claim as the unit of evaluation

2. Separating claims from evidence

3. Atomisation in pedagogy

4. Claims as semantic triples

5. Questions as atomic units

Why do we use it?

When do we use it?

How can I think about it?

Concepts to explore next

Check your understanding

Where this concept fits

Sources

Further reading

Footnotes

Graph View

Table of Contents

Backlinks