When people ask what knowledge is, they often picture a mental warehouse of facts. When they ask what intelligence is, they picture raw problem-solving power. Both images are useful, but they miss what is most important: knowledge and intelligence are not separate “things.” They are roles inside a larger system that turns experience into meaning, meaning into claims about the world, and claims into reliable action.
This article builds that system from the ground up, using the ideas you have been circling around: agents, propositions, concepts, relations, sharing across agents, and a category theory view of how knowledge becomes coherent.
1. Knowledge is not just information
Information is everywhere. A book, a website, a sensor, a database, even a rock with inscriptions. But none of that is knowledge by itself.
Knowledge appears only when an agent does at least four things:
- Perceives or receives signals from the world (or from other agents).
- Interprets those signals using concepts.
- Forms propositions (claims that can be true or false).
- Stabilizes those propositions through evidence, correction, and successful use.
So knowledge is not merely “data stored somewhere.” It is a pattern of meaning that survives contact with reality and remains useful under correction.
2. Intelligence is what moves through the space of knowledge
If knowledge is the stabilized content, intelligence is the capacity that builds, navigates, and reshapes that content.
Intelligence shows up as the ability to:
- Learn new concepts and refine old ones
- Form better propositions from messy input
- Discover relations that were not obvious before
- Update beliefs when evidence changes
- Transfer knowledge to new contexts
- Act effectively under uncertainty
A simple way to think about the relationship:
- Knowledge is the structure you have.
- Intelligence is your ability to grow, repair, and use that structure.
A person can have a large structure but weak ability to adapt it. Another person can have a smaller structure but strong ability to extend it. The strongest real-world performance usually comes from both.
3. The core unit: propositions
A proposition is the smallest unit that can directly count as knowledge.
A proposition is:
- About something
- Says something is the case
- Can be true or false
Examples:
- “The door is open.”
- “This chemical reaction releases heat.”
- “This market signal predicts demand next quarter.”
Questions, commands, and feelings are not propositions in this strict sense. “Is the door open?” is a question. “Open the door” is a command. “I like this song” expresses a preference. They can be meaningful, but they have a different logical role.
Why propositions matter so much:
- You can only test what has truth conditions.
- You can only debate what can be right or wrong.
- You can only correct what can fail.
So the moment your mind turns experience into a claim that could be mistaken, you are in the territory where knowledge can form.
4. Concepts: how an agent makes meaning
Concepts are the mental and linguistic “handles” that let an agent group the world into usable categories.
Examples:
- DOG, TREE, RED, HOT
- CONTRACT, RISK, SECURITY
- CAUSE, BEFORE, INSIDE
A key point: concepts are not true or false. They are tools for carving up experience. They decide what counts as “the same kind of thing” across different situations.
Concepts do three critical jobs for knowledge:
A. Compression
Without concepts, every moment is new. With concepts, many different experiences become one category. This makes memory and learning possible.
B. Recognition
Concepts let an agent pick out what matters: “That is a dog,” not just “a moving patch of color.”
C. Shared reference
Concepts are the bridge between minds. When two agents align on concepts, they can align on propositions built from them.
5. Relations: the glue that turns concepts into structure
Concepts alone are like nouns in a bag. Relations connect them into a model.
Examples of relations:
- “is a” (DOG is an ANIMAL)
- “has” (CAR has WHEELS)
- “causes” (FIRE causes HEAT)
- “located on” (CUP is on TABLE)
- “greater than” (A is larger than B)
Relations matter because knowledge is not a list. Knowledge is a web. The power of a knowledge system comes from how concepts are connected and how those connections support prediction and explanation.
6. Propositions are made from concepts plus relations
Most propositions have this form:
- Entity or concept + relation + entity or concept
Examples:
- “This apple is red.” (thing + property relation + property concept)
- “A causes B.” (event + causal relation + event)
- “Dogs are animals.” (concept + category relation + concept)
This is why concepts and relations are the real building blocks. Propositions are what you get when you assert that a relation holds.
And this is also where many disagreements come from:
- Two people may agree on the relation but disagree on the concept boundaries.
- Two people may share the words but attach different relations.
- Two people may share both but have different evidence about whether the proposition holds.
7. Why knowledge is tied to agents, yet still shared
If knowledge is “inside” agents, why does it not fracture into private worlds?
Because agents are connected by three stabilizers:
A. A shared reality that pushes back
Reality acts like a strict judge. Beliefs that do not match the world tend to fail when used for action and prediction. This pressure makes knowledge converge over time, especially in engineering and science.
B. Communication that transmits propositions
Language and symbols let agents export propositions. This is not automatic knowledge transfer, because the receiver still has to interpret the proposition through their own concepts and relations. But when alignment is good enough, propositions move between minds.
C. Social error correction
Groups often outperform individuals at correcting error, because disagreement forces justification. Institutions like science formalize this with replication, peer review, and measurement standards.
Shared knowledge is not identical knowledge. It is compatible knowledge that remains coherent under mutual checking.
8. Where intelligence lives in this pipeline
If you look at the full pipeline:
- signals → concepts → relations → propositions → tests → revision → action
Intelligence is the set of abilities that improves each step:
- Learning better concepts (cleaner categories, fewer distortions)
- Discovering better relations (especially causal and structural ones)
- Forming sharper propositions (more precise, more testable)
- Testing efficiently (good experiments, good questions)
- Revising gracefully (updating without collapsing)
- Acting robustly (generalizing beyond the training environment)
So intelligence is not separate from knowledge. It is the process that creates and maintains knowledge.
9. A category theory lens on knowledge
Category theory is not a definition of knowledge. It is a structural language for talking about systems made of “things” and “mappings,” and what stays invariant when you transform one structured system into another.
This turns out to match how knowledge behaves.
Concepts as “things”
In a category theory view, you can treat concepts like the fundamental items in a system. What matters is not what a concept “is made of,” but how it connects to other concepts.
Relations as “mappings”
Relations behave like mappings that link concepts. The emphasis is on composition: if one relation connects A to B, and another connects B to C, then there is a meaningful pathway from A to C.
This captures a basic feature of reasoning: knowledge grows through chained relations.
Propositions as structure that must cohere
In real knowledge, it is not enough to have random connections. The connections must fit together consistently. Category theory has a clean way to express this idea: different routes through the structure should agree when they are supposed to agree.
In plain terms: if you can reach the same conclusion by different legitimate paths, and those paths do not conflict, the knowledge web is coherent.
Sharing knowledge as “structure-preserving translation”
When one agent teaches another, the goal is not to copy raw experiences. The goal is to preserve structure:
- map the teacher’s concepts to the student’s concepts
- map the teacher’s relations to the student’s relations
- keep key patterns intact so that inferences still work
When translation fails, it is often because the structure is not preserved. The words may transfer, but the concept boundaries or relation meanings do not line up.
This lens also explains why some disagreements are hard: two agents may be operating with genuinely different conceptual structures, and there is no clean translation without changing one structure or the other.
10. A practical takeaway
If you want to build knowledge in yourself or in an organization, the most important work is often not collecting more facts. It is improving the underlying structure:
- Define concepts clearly and check that people mean the same thing by the same words.
- Make relations explicit, especially causal claims and dependency chains.
- Turn vague beliefs into testable propositions.
- Create feedback loops that punish error early and reward correction.
- Invest in translation between agents, including examples, counterexamples, and shared reference points.
When you do that, intelligence has something solid to operate on, and knowledge becomes something the group can actually share.