Representations, Schemas, Mental Models, and Explanatory Frameworks: What We Really Learn

Imagine two history books in front of you: one tells the legendary stories of Zhao Kuangyin conquering the world, while the other analyzes the patterns of political struggle in the Northern Song Dynasty. Do you know which one is more worth reading?

Some might argue that all knowledge is equal and both books are equally valuable. I strongly disagree. I believe the latter is far more deserving of your time and serious study.

We have previously discussed "Information Value": a piece of information is only valuable when it can change your actual actions. If even information isn’t equal, how can knowledge be?

Perhaps you read fifty books a year, keep meticulous notes, and collect every obscure factoid you encounter. You pride yourself on your breadth of knowledge, moving from geopolitics to quantum mechanics at dinner parties. While this "erudition" might provide mental pleasure, it is often practically useless.

When faced with real-world problems, you still don’t know how to make decisions. When reading the news, you can’t see the underlying patterns. If someone asks about the economy, you recite three conflicting expert opinions but dare not decide who is right. In this regard, you are less useful than an AI or even a search engine.

Most people are blinded by exam-oriented education, treating learning as the rote memorization of "knowledge points." They fail to understand the true value of learning or the methods to achieve it. This lesson introduces four conceptual tools: Representations, Schemas, Mental Models, and Explanatory Frameworks. These constitute the four levels of true learning.

Information vs. Knowledge: Learning is Compression #

First, let’s clarify why listening to historical stories—even true ones—is not "learning." Stories, tourist impressions, or twenty years of experience from a master craftsman provide "information," not "knowledge."

Information is the resolution of uncertainty. If you didn’t know whether there was oil under a patch of land and I tell you there is, that is information. It is valuable, but it isn’t knowledge because you haven’t learned a skill: given a different location, you still wouldn’t know how to judge its oil potential.

Information is "giving a man a fish," while knowledge is "teaching him how to fish." Knowledge is a universal operational method. To learn knowledge, we must ignore trivial, irrelevant details and extract the universal structure.

Professor Naftali Tishby of the Hebrew University of Jerusalem proposed the "Information Bottleneck Theory" around 2000 [1]. He argued that for both the human brain and deep neural networks, the essence of learning is not memory, but "compression."

Learning is compression. We seek patterns, not details. The more concise the pattern you take from a book, and the more complex phenomena it can explain, the higher your compression rate and the higher the quality of your knowledge. Patterns are higher-leverage than stories. Stories are merely the raw data used to feed the extraction of patterns.

Patterns are not one-dimensional; they have a layered structure. Let’s examine them from the lowest to the highest level.

Layer 1: Representation — Substitutes for Reality #

The first level is “Representation.” Representations are the substitutes in our minds for various things in the real world.

For example, the word "apple" is a representation. A real apple is a collection of atoms; the apple in your mind is something red, round, and sweet. Modern cognitive psychology suggests that our brains cannot process the real world directly—whenever we imagine or think about the world, we are manipulating representations [2].

In other words, for an object to cross the "Markov Blanket" and enter our consciousness, it must first be represented. You don’t think about actual economic activity; you think about economic concepts and indicators. You don’t grasp history itself, but rather a historical narrative. In our minds, we don’t manipulate territories; we manipulate maps.

Representations are the smallest units of our cognitive maps, corresponding to concepts, nouns, objects, relationships, and boundaries. Entering a new field means extracting these key representations: What are the core concepts? What are the most important variables? How do they relate?

Layer 2: Schema — Templates for Pattern Recognition #

The second level is "Schema." Simply put, if representations are individual cells, schemas are the tissues they form.

The concept was first proposed by British psychologist Frederic Bartlett in 1932 [3] and later popularized by Jean Piaget. If a representation is a single Lego brick, a schema is a recognizable structure built from those bricks—like a racing car chassis.

Schemas are templates for "pattern recognition." People with schemas in their minds see things in "chunks" rather than as isolated elements. When you hear "academic paper," you immediately expect sections like problem statement, methodology, results, and discussion. When you hear "startup," you think of value propositions, funding narratives, and organizational scaling.

Schemas allow you to quickly identify a situation because you can "fill in the blanks." Because we can abstract a schema from one narrative and apply it to another, we can learn by analogy. Once you understand the "platform economy" schema, you realize that Alibaba, Meituan, and Uber are essentially the same thing.

Schemas save cognitive bandwidth. This is why those with a solid foundation learn new things faster. Without schemas, every plot point is a surprise; with them, everything is a "trope."

Layer 3: Mental Models — The Logic of How Things Work #

The third level is "Mental Models." These are like cognitive organs because they are dynamic and have their own behavioral logic. A mental model is essentially a "working schema."

The term was coined by Scottish psychologist Kenneth Craik in 1943 [4]. It refers to a small-scale model of reality constructed in the brain to simulate how things work.

Mental models have specific functions, internal variables, causality, feedback loops, and boundary conditions. They allow for deduction: "If I do A, then B will happen." While a schema answers "What is this?", a mental model answers "How does it work?"

Mastering a mental model is the hallmark of truly learning a subject. For instance, the physicist Richard Feynman was famous for solving complex integrals using a technique called "differentiation under the integral sign" [5]. This was one of his mental models. Because he understood the underlying mechanism, he could apply it flexibly to various problems.

Charlie Munger famously advocated for a "lattice of mental models" covering multiple disciplines [7]. The fundamental purpose of study and research is to extract these models: What is the mechanism? What are the key variables? What are the feedback loops and side effects? Where does the model fail?

Layer 4: Explanatory Framework — Systematic Academic Insight #

The fourth and highest level is the "Explanatory Framework."

An explanatory framework is a "master map" of a field. It contains multiple mental models and schemas, provides the representations for all key questions, and explains how a broad class of phenomena should be understood.

For example, Professor Qin Hui’s Lectures on Qin and Han History provides an explanatory framework for "Imperial China." It doesn’t just list facts; it explains the underlying logic of its economy, politics, management, and culture. Once you grasp this framework, you understand why Imperial China had its specific fiscal and social structures.

When you reach the level of an explanatory framework, you begin to understand the world like a scholar: asking for mechanisms rather than stories, and weighing explanatory power rather than just taking a stance.

Summary: The Art of Compressing the World #

What, then, is the goal of learning? It is to compress the world into an internal map that can predict, explain, intervene, and self-correct:

• The materials of the map are "Representations."
• The quick templates are "Schemas."
• The working mechanisms are "Mental Models."
• The systematic understanding of the whole is the "Explanatory Framework."

Reading a book from "thick to thin" means discarding trivial information to extract these four structures. Reading it from "thin to thick" means using this internal map to handle any relevant scenario and generate new insights.

In the future, don’t ask "What is this article about?" Ask what structures you can take from it. Representations are just a glossary; schemas show you the patterns; mental models allow you to use the knowledge; and explanatory frameworks help you decide what is worth believing.

Learning in the AI Era: Mastering a Course in 48 Hours #

In March 2026, a report on X described an MIT graduate student who used Google’s AI application, NotebookLM, to compress a semester’s worth of learning into 48 hours [9].

He uploaded six versions of textbooks, 15 research papers, and all available lecture notes. He then asked the AI two critical questions:

1. What are the five core mental models that every expert in this field possesses?
2. What are the three biggest controversies in this field, and what are the strongest arguments for each side?

The first question sought mental models; the second mapped out the explanatory framework.

After grasping these, he had the AI generate ten questions to test him. He looked for the answers in the original texts, forcing himself to read with a purpose. After 48 hours, he was ready to engage in deep dialogue with his professor.

This student wasn’t just gaming an exam; he was truly learning. He understood that the essence of learning lies not in the time spent, but in capturing the core structures of knowledge.

Notes

[1] Tishby, Naftali, Fernando C. Pereira, and William Bialek. 1999/2000. “The Information Bottleneck Method.” [2] Pitt, David. 2023. “Mental Representation.” The Stanford Encyclopedia of Philosophy. [3] Bartlett, Frederic C. 1932. Remembering: A Study in Experimental and Social Psychology. [4] Craik, Kenneth J. W. 1943. The Nature of Explanation. [5] Goldmakher, Leo. 2021. “Differentiation Under the Integral Sign.” [6] Elite Daily Season 4, "Thinking Like a Rocket Scientist 2: Why Can Musk Use First Principles?" [7] Munger, Charlie. 1994/1995. “A Lesson on Elementary, Worldly Wisdom As It Relates to Investment Management & Business.” [8] Explanatory frameworks are similar to Thomas Kuhn’s "Paradigms" in The Structure of Scientific Revolutions. [9] Ali, Ihtesham. 2026. X post, March 7.