Cognition as Model Building and Revision
A unified model-based framework for essence, knowledge, learning, understanding, thinking, transfer, and problem solving.
Chinese original: 認知：建立、操作與更新模型.
Editorial note: Complete English translation of the user-supplied
knowledge.txt, received on September 28, 2026. The source’s definitions, examples, formulas, and claims are preserved as a working perspective, not independently verified scientific conclusions or evidence of learner mastery. Multiplicative formulas are conceptual illustrations, not validated quantitative measures. This article is distinct from Essence, Models, and Higher-Order Cognition and complements Learning from Models to Performance.
This entire set of questions can be connected through one unified framework:
Learning changes internal models; understanding builds models; abstraction compresses models; transfer reuses models; thinking operates on models; problem solving uses models to change reality; ability is the stable, effective functioning of this whole process; and seeing through to essence means identifying the few variables, relationships, mechanisms, and constraints most worth modeling.
Let us unpack this layer by layer.
1. What is “essence”? What is the essence of essence?
Essence is not “a mysterious truth hidden behind things.”
A more practical definition is:
Essence = the structures, relationships, mechanisms, and constraints that remain relatively stable beneath extensive surface variation and can efficiently explain, predict, and generate phenomena.
For example:
A table’s color, price, and material may change. But if your question is “Why can it support weight?”, what really matters is its load-bearing structure, not its color.
A company’s logo, office, and slogan may change. If the question is “Why does it remain profitable over the long term?”, what matters more may be:
Customer needs → value creation → transactions → costs → competitive advantage → cash flow.
So:
Appearance answers “What does it look like?”;
regularities answer “What usually happens?”;
mechanisms answer “Why does it happen this way?”;
essence answers “What, if changed, would fundamentally change the entire phenomenon?”
The essence of essence
This can be condensed into four terms:
Invariants, causality, constraints, and generation.
In other words:
- What remains relatively unchanged amid change?
- Which factors actually cause the outcome?
- Which conditions limit what can happen?
- What simple mechanism can generate many complex phenomena?
One level deeper:
Seeing essence is essentially a search for the “smallest sufficient model.”
It is not about knowing the most, but finding the smallest set of key variables and relationships that can explain the most phenomena.
But there is a very important limitation:
A thing does not necessarily have just one unique “essence.”
Different questions and levels of analysis lead to different essences.
For example, “What is the essence of human beings?”
Biology, psychology, economics, sociology, and philosophy may give different answers.
The more precise question is therefore not:
“What is its one true essence?”
It is:
“For the phenomenon I want to explain, which factors are most central, most stable, and most causally influential?”
2. What is knowledge? What is its essence?
Many people treat:
Knowledge = remembered information.
But that is not really the case.
Information might simply be:
“Water boils at about 100°C near standard atmospheric pressure.”
For it to become knowledge means that it has entered your cognitive structure.
You know:
- What it is;
- why;
- what it relates to;
- when it holds;
- when it does not;
- how to use it;
- how to derive it from other things.
Therefore:
Knowledge = organized internal models that can be retrieved and used for explanation, prediction, and action.
The essence of knowledge is not “the amount of information.”
It is:
Structured, usable models.
Two people reading the same book may therefore acquire completely different knowledge.
One acquires 100 sentences.
The other acquires:
5 concepts + 3 relationships + 2 causal mechanisms.
The second person’s knowledge is usually stronger.
3. What is learning? What is its essence?
Learning is not:
Reading books, attending classes, or taking notes.
These are only ways of supplying input for learning.
Actual learning is:
Experience producing a relatively lasting change in your internal models or behavioral strategies.
It can be represented as:
Old model
→ encounter reality
→ make a prediction
→ discover an error
→ revise the model
→ test again.
Therefore:
The essence of learning = updating models through feedback.
If you read for 10 hours, but:
- Your way of judging has not changed;
- your predictive ability has not changed;
- your way of solving problems has not changed;
- your way of acting has not changed;
then very little actual learning may have occurred.
4. What is understanding? What is its essence?
“I understood what I read” is one of the easiest statements to turn into an illusion.
Genuine understanding means at least that you can:
Explain it, derive it, predict with it, analyze it under changed conditions, and use it to solve new problems.
For example, genuinely understanding supply and demand is not being able to recite:
When prices rise, demand falls.
It is being able to identify which factors are changing when you encounter:
Rent, concert tickets, wages, chips, taxis, luxury goods, and so on.
So:
Understanding = building a model in your mind that can run.
And:
The essence of understanding = grasping internal relationships, causality, structure, and boundaries.
To judge whether you truly understand, ask:
If the conditions change, what will happen to the outcome? Why?
If you cannot answer, you usually have not genuinely understood yet.
5. What is abstraction? What is its essence?
Suppose you see:
An apple falling, a ball falling, and a stone falling.
At an elementary level, cognition sees:
Three different events.
After abstraction, it sees:
Objects accelerate under gravity.
So:
Abstraction = ignoring unimportant differences while retaining important common structures.
Its core is:
Remove details; preserve invariants.
But abstraction is not simply “becoming vague.”
Good abstraction is actually more precise.
For example:
Dogs, cats, humans, whales
↓
Mammals
One level higher:
Life
Higher still:
Self-organizing systems.
Therefore:
The essence of abstraction = finding structures that remain stable across cases.
It can also be understood as:
Cognitive compression.
100 examples → 1 regularity.
6. What is transfer? What is its essence?
You learn how to handle one problem, then encounter another that looks completely different, yet can still use the original method.
That is transfer.
For example:
You learn in a board game that “local gains must not sacrifice the overall position.”
Later, you may find that this also applies to:
Business, negotiation, career choices, and investment.
Not because they look alike on the surface.
But because:
Their underlying structures are similar.
Therefore:
Transfer = mapping a structure learned in situation A onto situation B.
Its essence is:
Recognizing structural isomorphism.
The stronger your ability to abstract, the easier transfer usually becomes.
Because:
Those who look at appearances see differences;
those who look at structure see similarities.
7. What is ability? What is its essence?
Ability is not “knowing.”
Nor is it succeeding occasionally.
A better definition is:
Ability = an internal system that reliably produces a particular class of effective results under different conditions.
For example, being able to swim is not knowing swimming theory.
It is your body still being able to perform the relevant actions in different water conditions.
Ability therefore usually includes:
Knowledge + representation + skills + strategies + judgment + feedback-based correction.
So:
The essence of ability = reliably transforming input into effective output.
8. What is thinking? What is its essence?
Thinking is not “having many voices in your head.”
It is closer to:
Operating on internal representations.
For example, you think:
“What would happen if the price increased by 20%?”
It has not happened in reality yet.
But the mind is simulating it.
So:
Thinking = the process of building, transforming, comparing, combining, searching, and evaluating internal models.
There are not that many core operations of thinking:
Classification, comparison, causal reasoning, analogy, decomposition, combination, deduction, induction, counterfactual reasoning, and simulation.
Therefore:
The essence of thinking = low-cost trial and error in model space before acting in reality.
9. What is a problem? What is its essence?
A problem is not:
“Something bad is happening.”
A more precise definition is:
There is a gap between the current state and the goal state, and how to bridge it is not fully known.
A complete problem therefore includes at least:
Current state S
Goal state G
Constraints C
Available actions A
An unknown path.
If the path is fully known, it is not really much of a problem.
It is simply an execution task.
So:
The essence of a problem = finding a feasible path from the present to the goal under constraints.
10. What is learning ability? What is its essence?
Learning ability is not memory.
It can be approximately understood as:
The amount of effective model updating achieved per unit of time and experience.
Strong learners can usually complete this cycle more quickly:
Acquire → model → test → discover errors → revise → abstract → transfer.
So:
The essence of learning ability = updating your internal models with high quality and efficiency.
The most important way to improve learning ability is not to increase “study time,” but to improve the quality of this cycle:
Input → recall → output → feedback → correction → output again.
Some of the most effective habits are:
- Reread less; actively recall more. Close the book and explain it yourself.
- Copy answers less; generate answers more. Guess, reason, and try first.
- Get feedback quickly. Do not let errors accumulate for days.
- Build relationships between pieces of knowledge. Ask not only “What is A?” but also “How does A relate to B?”
- Use spaced review rather than concentrated binge-reading.
- Interleave practice. Do not always practice only one kind of problem.
- Find a new situation in which to use what you learned. This trains transfer.
11. What is the ability to understand? How can it improve?
The ability to understand can be viewed as:
The capacity to recover internal structure from information.
Someone gives you a sentence.
Weak understanding:
Remembering the sentence.
Strong understanding:
Seeing the concepts, relationships, premises, causality, intentions, and limitations behind it.
To improve understanding, you can repeatedly ask about anything:
What is it?
Why?
How does it happen?
What conditions does it depend on?
What is it similar to?
What is it different from?
If X changes, what happens to Y?
Can I give a counterexample?
Can I derive it myself?
One particularly powerful exercise is:
After reading a passage, reconstruct it in your own words without looking at the original.
Do not aim to repeat the original sentences.
Aim to:
Reconstruct its logic.
12. How can we improve the ability to understand regularities?
A regularity is not:
“A often appears together with B.”
What you really need to understand is:
Under what conditions, and through what mechanism, does A cause B?
When studying regularities, consistently ask five things:
What are the variables?
How are they related?
What is the underlying mechanism?
Under what conditions does it hold?
Under what conditions does it fail?
For example:
“Effort → success”
This is not a good regularity.
It leaves out many variables.
A better model might be:
Ability × effort × direction × resources × opportunity × feedback
And many of these do not combine through simple addition.
People who genuinely understand regularities actively look for:
Counterexamples.
Because counterexamples expose a regularity’s boundaries.
13. How can we better understand what individuals and groups “say, do, and think”?
The most important point here is:
What people “think” cannot be observed directly; it can only be inferred probabilistically from evidence.
Do not treat your guesses as mind-reading.
To understand someone’s behavior, consider at least these things together:
What they said
What they did
What they gained
What they fear losing
What constraints they face
What role they occupy
Whom they are dealing with
How they have behaved in the past.
A useful framework is:
Behavior ≈ goals × incentives × information × ability × constraints × social norms.
For example, an employee says:
“This project is excellent.”
You cannot directly conclude:
“They genuinely think it is excellent.”
Other possibilities include:
- They genuinely agree;
- they do not want to clash with their manager;
- they lack sufficient information;
- they see no benefit in opposing it;
- they are maintaining group relationships.
To understand people, therefore, do not listen only to words.
Look at these together:
Language, behavior, costs, incentives, roles, and constraints.
Understanding groups also requires:
Power structures, identity, norms, common knowledge, distribution of interests, and coordination mechanisms.
Many groups that appear to have “everyone in agreement” may actually be situations where:
Each person merely believes that everyone else agrees.
14. What is the ability to abstract? How can it improve?
The ability to abstract is:
The capacity to extract common structures from many different cases.
One particularly powerful training method:
For everything you learn, find 3 examples that look completely different on the surface.
For example, you learn:
“The bottleneck determines system throughput.”
Look for examples in:
Factories, traffic, computers, company management, and learning.
Then ask:
They look completely different, so why can the same model apply to all of them?
A second method is called:
Expressing a case in terms of variables.
For example:
“Xiaoming reads for 2 hours every day, so his grades improve.”
Do not stop at the person’s name and the numbers.
Abstract it into:
Time invested → amount of effective practice → change in ability → change in performance.
Go further:
X → mediating mechanism M → Y.
Your thinking then moves from a story to a model.
15. What is problem solving?
Problem solving is not:
Finding an answer.
It is:
Finding and executing an effective path from the current state to the goal state.
Its basic cycle is:
Define the problem
→ model it
→ find the bottleneck
→ generate options
→ select an option
→ act
→ observe the result
→ update the model.
So:
The essence of problem solving = model-guided search and trial and error.
Many people cannot solve a problem not because they are “unintelligent,” but because they represented it incorrectly from the outset.
An idea often paraphrased and attributed to Einstein is important here:
How a problem is represented often determines how difficult it is to solve.
16. What is problem-solving ability?
It can be roughly expressed as:
Problem-solving ability ≈ problem representation × knowledge × reasoning × search strategy × feedback × execution.
The most easily overlooked component is:
Problem representation.
For example:
“How can I become more self-disciplined?”
This may be a poorly framed problem.
Represent it differently:
“Which environmental conditions make me likely to scroll on my phone after 10 p.m.?”
It immediately becomes actionable.
Another example:
“Why is the company not making money?”
Break it down into:
Traffic × conversion rate × average transaction value × repeat-purchase rate − costs.
Once the representation changes, the space of possible solutions changes too.
17. How can problem-solving ability improve?
When encountering a problem, do not immediately search for an answer.
First, consistently follow this procedure:
First layer: Definition
What result do I actually want?
Second layer: Current situation
What is actually happening now?
Do not treat guesses as facts.
Third layer: Decomposition
Which key variables determine the outcome?
Fourth layer: Bottleneck
Which factor currently limits the outcome most?
Fifth layer: Mechanism
Why has this bottleneck formed?
Sixth layer: Options
What actions could change it?
Seventh layer: Minimal experiment
How can I test my judgment at the lowest cost?
Eighth layer: Feedback
Does the result match the prediction?
If it does not:
Do not merely work harder; update the model first.
This is very important.
18. How can we improve the ability to “see through to essence”?
You can turn it into a consistent algorithm.
For any complex phenomenon, ask:
1. What is merely appearance?
Names, forms, stories, emotions, and packaging.
2. Which variables really matter?
Do not list 50.
Try to identify 3–7.
3. What remains unchanged across different cases?
This is a search for structure.
4. Which factors actually cause which outcomes?
This is a search for causality.
5. What constraints affect the system?
Time, resources, information, physical conditions, institutions, and incentives.
6. Where is the biggest bottleneck?
The outcomes of many complex systems are mainly limited by a small number of bottlenecks.
7. What simple model can generate many phenomena?
This is compression.
8. If my model is true, what else should it predict?
This is verification.
9. What counterexamples exist?
This is a search for boundaries.
10. If one variable changes, will the entire outcome change?
This identifies core variables.
19. Condensing all your questions into a “cognitive map”
You can view the entire process as:
Reality
↓
Perception and information
↓
Knowledge
Form usable internal models
↓
Understanding
Know how those models work internally
↓
Abstraction
Remove details and retain structure
↓
Transfer
Apply the same structure to a new situation
↓
Thinking
Operate on, combine, and simulate these models
↓
Problem solving
Use models to find a path from the current state to the goal
↓
Ability
The processes above become stable, efficient, and automatic
Running throughout the whole process is:
Feedback.
If I had to answer in a single sentence:
How do people become increasingly “intelligent”?
Not simply by remembering more.
But by:
Building better models, discovering their errors more quickly, correcting them more quickly, and transferring them to new problems.
20. One final layer deeper: What is the shared “essence” of these abilities?
You asked about:
Learning ability, understanding, abstraction, transfer, problem solving, and seeing essence.
They look like many separate abilities.
But they are highly related underneath.
I would condense them into four core capacities:
Representation, compression, inference, and updating.
Representation:
Turn reality into a good model.
Compression:
Extract a few regularities from many phenomena.
Inference:
Use a model to derive outcomes not yet observed.
Updating:
Revise a model when reality does not match it.
Thus:
Learning = updating models.
Understanding = building runnable models.
Abstraction = compressing models.
Transfer = reusing models across situations.
Thinking = operating on models.
Problem solving = using models to search for actions.
Ability = running models reliably.
Seeing essence = finding the smallest model with explanatory power.
If you truly want to improve this whole set of abilities systematically, you do not need to train more than a dozen separate systems.
Choose one real problem each day and repeatedly practice:
Observe phenomena → distinguish facts from explanations → identify variables → identify relationships → identify causality → identify constraints → identify invariants → build a model → make predictions → find counterexamples → test in practice → revise the model through feedback → find another situation for transfer.
Over the long term, this may be the most unified route for training learning, understanding, abstraction, thinking, problem solving, and seeing through to essence.