Essence, Models, and Higher-Order Cognition
A model-based perspective connecting essence, knowledge, learning, understanding, abstraction, transfer, thinking, and practical ability.
Chinese original: Essence, Models, and Higher-Order Cognition.
Editorial note: Full English translation of the user-supplied
knowledge.txt, added on September 27, 2026. The definitions, formulas, claims, and examples below preserve the author’s working perspective; they have not been independently fact-checked and are not evidence of the learner’s mastery. Formulas are conceptual illustrations, not validated quantitative measures.
What is essence, and what is the essence of essence?
What is knowledge, and what is the essence of knowledge?
What is learning, and what is the essence of learning?
What is understanding, and what is the essence of understanding?
What is abstraction, and what is the essence of abstraction?
What is transfer, and what is the essence of transfer?
What is ability, and what is the essence of ability?
What is thinking, and what is the essence of thinking?
What is learning ability, what is its essence, and how can it be improved?
What is the ability to understand, what is its essence, and how can it be improved? How can we improve our ability to understand regularities, and to understand what other people and groups say, do, and think?
What is the ability to abstract, what is its essence, and how can it be improved?
How can we improve our ability to see through to the essence of things?
Higher-order human cognition is essentially about finding stable structures in a messy, concrete world; building models that compress information, support prediction and action, and transfer to other settings; and continuously revising those models through feedback from reality.
From this perspective, knowledge, learning, understanding, abstraction, transfer, ability, and thinking are not seven unrelated things. They form a chain:
Phenomena → notice differences → extract information → build knowledge → develop understanding → abstract structures → transfer to new situations → reliably solve problems → ability.
And thinking runs through the entire chain.
1. What is essence?
Let me begin with a practical working definition:
Essence is the relatively stable core structure and mechanism that can generate, explain, or constrain many surface phenomena at the scale of a particular problem.
Essence is therefore not a sentence that merely sounds profound.
A genuine essence usually has four characteristics:
Stability, generativity, compression, and boundaries.
In other words:
It remains when the phenomena change; it can explain why many phenomena occur; a few principles can compress many facts; and you know under what conditions it holds.
For example:
An apple falling to the ground, raindrops falling, and a thrown ball falling are different phenomena.
Simply remembering:
Apples fall.
is a fact.
Discovering:
Many objects accelerate toward the ground.
is a regularity.
Then building a model:
Objects exert gravitational forces on one another, which can be described with an appropriate model at a particular scale.
brings us closer to mechanism and essence.
Then what is the essence of essence?
Going one level deeper:
The essence of essence is the stable constraints and relationships that generate phenomena.
We can even compress this into two terms:
Invariance + generativity.
The most important question in “seeing the essence” is therefore not:
What is this thing really?
It is:
Behind many different phenomena, what remains unchanged? What structure repeatedly generates these phenomena?
But note:
There is no single essence independent of the question, scale, and purpose.
For example, “What is water?”
At the chemical scale:
H₂O.
At the thermodynamic scale:
A substance with particular phase transitions, heat capacity, and other properties.
At the biological scale:
An important medium for life’s metabolic processes.
A sophisticated ability to “see through to essence” is therefore not an endless search for a single ultimate definition. It is finding:
The level of structure with the greatest explanatory power for the problem at hand.
2. What are knowledge, learning, understanding, abstraction, transfer, ability, and thinking?
We can view them together:
| Concept | What it is | Its essence at a deeper level |
|---|---|---|
| Knowledge | Representations of things, relationships, regularities, and methods in the world | Structured models that can be called upon |
| Learning | A lasting change in internal models or behavior through experience | Updating models based on error |
| Understanding | Being able to explain, derive, predict, and reconstruct something | Forming a generative model |
| Abstraction | Ignoring unimportant differences while retaining common structure | Extracting invariants |
| Transfer | Applying existing knowledge in a new situation | Recognizing the same underlying structure and remapping it |
| Ability | Reliably completing a class of tasks in different situations | Models, procedures, and control systems that can be repeatedly called upon |
| Thinking | Comparing, combining, reasoning with, and searching internal representations | Operating on models under goals and constraints |
These seven concepts have deep relationships with one another.
3. What is knowledge? What is its essence?
Information is not the same as knowledge.
For example:
The speed of light is approximately 299,792,458 m/s.
This is a piece of information.
If you can only recite it, it remains mainly something held in memory.
If you know its relationships with electromagnetism, relativity, time, space, and causal limits, it becomes part of your knowledge network.
So:
Knowledge = information that has been organized and can be called upon for explanation, prediction, or action.
What truly matters about knowledge is not its quantity, but its structure.
Low-quality knowledge looks like:
1
2
3
4
5
A
B
C
D
E
Five isolated memories.
High-quality knowledge looks like:
1
2
3
4
A → B
A → C
B + C → D
D, under certain conditions → E
Therefore:
The essence of knowledge is the usable models of the world in your mind.
4. What is learning? What is its essence?
Many people treat:
Reading, listening to lectures, highlighting, and copying notes
as learning.
But these are only input activities.
Genuine learning must change your internal system.
Therefore:
Learning = a relatively lasting change in your knowledge structures, ways of predicting, or behavioral strategies caused by experience.
One core cycle is:
Predict → act → obtain a result → detect an error → revise the model → predict again.
So:
The essence of learning is updating models.
If, after reading a book:
You still cannot solve problems you previously could not solve; you still cannot explain what you previously could not explain; you cannot handle a differently worded question; and a few days later you cannot recall any of it;
then you have usually only “encountered information” and have not yet completed deeper learning.
5. What is understanding? What is its essence?
“I can follow what I am reading” often does not amount to understanding.
A strong criterion for genuine understanding is:
Without the original text, can you generate it again yourself?
Suppose you say you understand supply and demand.
After genuinely understanding it, you should be able to:
Explain why prices change; predict what happens when a condition changes; analyze a new market; assess an exception; and explain when the model might fail.
So:
Understanding is not remembering answers; it is building a model in your mind that can generate answers.
We can express this as:
Understanding ≈ structure + causality/mechanism + boundary conditions + counterfactual reasoning.
Counterfactuals are particularly important:
What happens if X is removed?
What if X doubles?
What if condition A does not hold?
Someone who can only repeat the material is usually relying on memory.
Someone who can still derive results after conditions change is closer to understanding.
Therefore:
The essence of understanding is building a model that can compress phenomena and generate explanations and predictions.
6. What is abstraction? What is its essence?
Suppose you observe:
A company managing employees; an operating system managing a CPU; a city managing traffic; and a human body regulating temperature.
On the surface, these are completely different.
At a higher level of abstraction, however, you might find:
Limited resources → competing demands → allocation → feedback → adjustment.
This is abstraction.
Abstraction is not about “speaking mysteriously.”
Quite the opposite.
Abstraction is deliberately discarding irrelevant differences and retaining only the common structure relevant to the question.
Therefore:
The essence of abstraction = extracting invariants.
You can abstract three things:
Apples, oranges, and bananas
into:
Fruit.
Abstract further:
Food.
Further still:
Resources.
But good abstraction is more than classification.
The most valuable abstractions are often relational abstractions.
For example:
Teacher → student
Coach → athlete
Supervisor → employee
The roles look different, but may share:
An information gap → guidance → practice → feedback → a change in ability
This relational structure has more potential for transfer than simply classifying nouns.
7. What is transfer? What is its essence?
Transfer means:
Recognizing the underlying structure of an old problem in a new one and remapping an existing model onto it.
For example, if you genuinely understand negative feedback:
A thermostat is one example;
human temperature regulation is another;
company inventory control is another;
and machine control is another.
You have not memorized four separate bodies of knowledge.
Instead, you have mastered this structure:
Target value → measurement → deviation → adjustment → new state → measurement again
Therefore:
The essence of transfer is structural matching.
This also explains why:
Without abstraction, transfer is difficult.
If you remember only concrete appearances, you cannot recognize the same problem in a different wrapper.
8. What is ability? What is its essence?
Knowing does not mean doing.
Understanding does not necessarily mean doing either.
For instance, you may understand the principles of fitness, swimming, writing, or negotiation very well.
But actually performing them still requires extensive practice.
So:
Ability is the capacity to reliably produce effective results under particular constraints and variations.
It usually includes:
Knowledge models + operational procedures + attentional control + feedback-based correction + patterns from experience + situational judgment.
Therefore:
The essence of ability is not how much you know, but whether you can reliably turn models into results.
Genuine ability has another characteristic:
Robustness to change.
Only being able to solve problems you have already done is not particularly strong ability.
Still being able to solve them after the numbers change is stronger.
Recognizing the problem when the situation changes is stronger still.
Handling real situations with incomplete information comes closer to genuine ability.
9. What is thinking? What is its essence?
We can understand thinking as:
The mind manipulating possibilities within internal models without having to act on reality immediately.
Examples include:
- Comparing;
- classifying;
- decomposing;
- causal reasoning;
- hypothesizing;
- simulating;
- counterfactual reasoning;
- drawing analogies;
- abstracting;
- finding contradictions;
- making decisions.
Essentially, all of these operate on internal representations.
Therefore:
The essence of thinking is searching, combining, comparing, and reasoning with internal models under goals and constraints.
Why can thinking save enormous costs?
Because you do not have to try something a hundred times in reality.
You can first simulate it eighty times in your mind, then test it a few times in reality.
10. What is learning ability?
Learning ability is not “having a good memory.”
Nor is it “reading quickly.”
I would define it as:
A person’s ability to efficiently build, revise, retain, and transfer models from limited experience.
A rough expression is:
Learning ability ≈ acquiring useful information × modeling × feedback-based correction × memory retrieval × transfer × metacognition.
If any link is extremely weak, it limits the whole process.
For example, someone reads a great deal but never tests themselves:
Strong input, weak correction.
Someone has an excellent memory but does not abstract:
Strong retention, weak transfer.
Someone is very intelligent but does not know what they fail to understand:
Potentially strong models, weak metacognition.
11. How can learning ability genuinely improve?
The most effective approach is not to “learn more,” but to change your learning loop.
You can use the following training process over the long term:
- Predict first, then learn. Try answering before looking at the answer. Generating a prediction makes it easier for the mind to produce a genuinely useful prediction error.
- Reconstruct immediately after learning, with the material closed. Without looking, write down the core concepts, relationships, and chain of reasoning. What you cannot retrieve marks what you have not actually mastered.
- Represent knowledge as relationships. Do not only ask “What is A?” Ask what A affects, what affects A, and how A relates to B.
- Always practice variations. Change the numbers, background, wording, or domain while retaining the same principle. Variation forces you beyond surface patterns.
- Deliberately seek counterexamples. When does a concept fail to apply? Counterexamples force you to discover its real boundaries.
- Make predictions rather than after-the-fact explanations. Write down your judgment before the result appears. Otherwise, it is easy to develop the illusion that “I knew it all along.”
- Teach someone else. If you cannot reconstruct an idea in your own words, you usually have not formed a stable model yet.
- Review errors regularly. Do not merely record “the answer was wrong.” Record: “What incorrect model was I using? Why did it seem reasonable? What exactly changed in the new model?”
What most distinguishes highly effective learners is not that they never make mistakes.
Instead, they move quickly through:
Make an error → identify the incorrect model → update the model.
12. What is the ability to understand?
The ability to understand is the capacity to build high-quality internal models from limited information.
Given the same material, A sees ten isolated facts.
B sees:
Three variables, two causal chains, one feedback loop, two assumptions, and one boundary condition.
Naturally, B “understands more deeply.”
The most important way to improve understanding is therefore not repeated reading.
It is converting any material into:
Elements → relationships → mechanisms → conditions → outcomes.
When encountering something, ask:
“What are the core variables?”
Then:
“What affects what?”
Then:
“Through what mechanism?”
Then:
“Under what conditions?”
Finally:
“If one variable changes, what happens to the outcome?”
When you can begin answering the final question, your understanding is usually relatively deep.
13. How can we improve our ability to understand regularities?
This is particularly important.
People easily mistake correlation for a regularity, then mistake a regularity for causation.
For example:
A often occurs together with B.
That does not mean:
A causes B.
It could be:
A → B
Or:
B → A
Or:
1
2
C → A
C → B
It could even be coincidence.
For any supposed regularity, therefore, identify at least:
Variables, relationships, mechanisms, and boundaries.
I would define a regularity as:
A stable relationship or generative mechanism that repeatedly appears among certain variables under particular conditions.
When understanding regularities, do not only ask:
What happened?
Ask:
Why does it keep happening?
For example, a problem keeps recurring in a company.
A surface observation:
A particular employee made a mistake.
Deeper:
There is a gap in the process.
Deeper still:
The KPIs encouraged this behavior.
Deeper still:
The information structure, incentive structure, and allocation of authority and responsibility jointly generated this outcome.
This is moving down through:
Events → patterns → structures → mechanisms.
Usually, the further you go, the closer you come to essence.
14. How can we better understand what a person says, does, and thinks?
There is an extremely important principle here:
Do not turn understanding people into mind-reading.
You can never directly observe another person’s inner world.
You can only build probabilistic models from evidence.
When understanding someone, separate these kinds of information:
What they say; what they do; what costs they bear; what benefits they receive; what information they possess; what role they occupy; and how they have repeatedly acted in the past.
In particular:
Words are evidence, but not all the evidence.
For example, someone says:
“This is not important to me.”
But they continue investing substantial time, money, and emotion in it.
You should at least notice:
There is a contradiction between the statement and the behavior.
But do not immediately jump to:
“So they must be lying.”
There may be other explanations.
Strong interpersonal understanding means:
Holding several hypotheses at once, then eliminating them using subsequent evidence.
Not:
First impression → immediate judgment of motive.
15. How can we understand what groups say, do, and think?
Groups are more complex than individuals.
Because:
Group behavior is not simply the sum of everyone’s thoughts.
For example, a company’s final decision may produce an outcome that no individual truly wanted.
But:
The performance system works that way; information becomes distorted as it moves upward; each department protects itself; the accountability system encourages risk avoidance; and everyone waits for everyone else.
The eventual outcome is one that “nobody truly wanted, but everyone jointly caused.”
To understand groups, pay particular attention to:
Incentive structures, power structures, information flows, rules, identities, norms, networks, coordination methods, and feedback loops.
This is an important cognitive advance:
Do not explain groups only in terms of what kind of people their members are.
Also ask:
What structure would lead many different people, after entering this system, to gradually develop similar behaviors?
That question usually takes you deeper.
16. What is the ability to abstract, and how can it improve?
The ability to abstract can be defined as:
The capacity to identify common structures across different concrete cases and form concise representations of them.
One powerful way to improve it is:
Comparing multiple examples.
Do not study only one example.
For instance, when learning about feedback systems, study:
Thermostats, human blood glucose, business inventories, social media recommendations, and financial markets.
Then ask:
How are they completely different?
And, more importantly:
What relationships do they share?
You gradually see:
Input → system → output → measurement → adjustment.
This is abstraction.
Another effective exercise is:
Three examples and one counterexample.
For every concept you learn:
Create three very different examples, then create one apparent example that looks similar but actually does not fit.
For instance, monopoly.
If all you can name is:
Google, a particular telecommunications company, and so on,
you may merely be remembering examples.
Only when you can judge:
What constitutes a monopoly, what does not, and where the boundaries lie,
do you begin genuinely forming the concept.
17. A deeper secret of abstraction
Higher abstraction is not always better.
Many people get stuck in statements such as:
The world is all systems.
Everything is energy.
Everything is a game.
Everything is information.
These may sound profound, yet often have little practical explanatory power.
High-level abstraction requires two simultaneous abilities:
Abstracting upward + making things concrete downward.
In other words:
1
2
3
4
5
6
7
Concrete cases
↓
Discover a common structure
↓
Form an abstract model
↓
Use that model to explain new cases
The sign of truly mastering an abstract concept is not:
Being able to describe it very abstractly.
It is:
Being able to move freely up and down between the concrete and the abstract.
18. How can we improve the ability to see through to essence?
We can now give a complete answer.
Seeing through to essence is not a mysterious intuition.
It mainly combines:
Decomposition + abstraction + causal reasoning + systems thinking + counterfactual thinking + model comparison + awareness of boundaries.
For any complex problem, you can use eight questions about essence:
What is the phenomenon? → What keeps recurring? → What details are merely incidental? → What are the key variables? → How do they interact? → What mechanism generates the outcome? → Under what conditions does it hold? → Would the phenomenon still occur if a particular factor were removed?
Finally, ask an even more demanding question:
Can fewer principles explain more phenomena?
If so, you are usually approaching a deeper level.
19. Seeing through to essence often means descending through five levels
For example, a company’s performance is declining.
First level: event
Revenue fell this quarter.
Second level: pattern
It has fallen for several consecutive quarters.
Third level: variables
Customer acquisition is declining, churn is rising, and average transaction value is falling.
Fourth level: structure
Acquisition costs are rising while product retention is falling, yet the company continues relying on more advertising to maintain growth.
Fifth level: generative mechanism
The business model’s growth depends on continuously purchasing external traffic, while its unit economics have deteriorated.
Only now do we begin approaching:
Why do those surface problems keep recurring?
Seeing essence means refusing to stop at describing outcomes and tracing them back to the structures that generate them.
20. Compressing all these concepts into a set of relationships
Finally, let me compress the whole question:
Knowledge: What models you possess.
Learning: How you change models.
Understanding: Whether your models can generate explanations and predictions.
Abstraction: Whether you can extract common structures from many models.
Transfer: Whether you can recognize and use those structures in new situations.
Ability: Whether you can reliably use models to produce results.
Thinking: How you operate on, compare, search, and revise models.
Therefore:
Learning ability = efficiency in building and updating models.
The ability to understand = the capacity to build generative models.
The ability to abstract = the capacity to identify structures that remain invariant across situations.
The ability to transfer = the capacity to recognize and use abstract structures again in new situations.
And:
The ability to see through to essence = the capacity to look past surface differences and identify the core structures, constraints, mechanisms, and boundaries that generate phenomena.
If you remember only one overall formula
I would give you this one:
Phenomena → patterns → variables → relationships → structures → mechanisms → boundaries → predictions → verification → transfer
Repeatedly following this path of thinking will improve your understanding, abstraction, learning, and insight together.
At a deeper level:
Truly intelligent people do not necessarily know the most.
Often, when looking at the same hundred phenomena, others retain a hundred pieces of information, while such a person retains only a few highly compressive models with strong explanatory power and transferability.
Those few models can then regenerate the hundred phenomena.
This is a central difference between understanding and merely knowing.