
How to ask Super Intelligence to explain difficult ideas at different levels is not simply asking for “ELI5” and then asking for a more advanced version. Good explanation changes the representation, vocabulary, examples and formal detail while preserving the underlying mechanism.
A simplified explanation becomes dangerous when it teaches a rule that stops being true as soon as the learner advances. An advanced explanation becomes useless when it assumes prerequisites the learner does not have.
In the eduKateSG life series, Super Intelligence, or SI, is our editorial name for practical AI assistance. This article follows How to Build a Personal Super Intelligence Tutor and focuses specifically on explanation calibration.
The central rule is: simplify the access path, not the truth of the idea.
The Explanation Problem
When learners say an idea is “too hard”, several different problems may be present.
- missing vocabulary;
- missing prerequisite concept;
- too much abstraction;
- too many steps at once;
- an example that does not map cleanly to the mechanism;
- formal notation before intuition;
- intuition without formal precision;
- explanation at the wrong output level.
SI can generate explanations instantly, but it needs to know which problem it is solving.
The Explanation Ladder
Use five levels:
Level 1 — Concrete
One tangible example with minimal vocabulary.
Level 2 — Intuitive
The main relationship in ordinary language.
Level 3 — Operational
What to do, how the parts interact and where the idea applies.
Level 4 — Formal
Precise definitions, notation, equations, rules or mechanisms.
Level 5 — Transfer and limits
Boundary cases, exceptions, neighbouring concepts and application in unfamiliar contexts.
The learner does not always need all five levels in one session.
Start from the Learner’s Current Representation
Before asking SI to explain, ask the learner to describe what they currently think.
“Explain photosynthesis” is less diagnostic than “I know plants use light and carbon dioxide, but I do not understand where the new plant mass comes from.”
“Explain functions in coding” is less diagnostic than “I can write repeated code, but I do not understand why a function returns a value instead of just printing it.”
The second form tells SI which bridge is missing.
The Prerequisite Check
An explanation cannot substitute for a missing prerequisite forever.
Ask SI: “Before explaining this, list the minimum prerequisite ideas. Give me one quick check for each. If a prerequisite is unstable, explain that first.”
This prevents advanced explanations from becoming a vocabulary fog.
The Three Axes of Explanation Level
Vocabulary level
How much technical language can the learner use?
Abstraction level
Can the learner reason with symbols and models, or do they need concrete cases first?
Mechanism depth
Does the task require intuitive use, operational procedure or causal/formal understanding?
A learner can be advanced on one axis and beginner on another. A programmer may understand abstract logic but be new to a particular framework’s vocabulary.
Age Is Not the Same as Level
Do not equate “child” with “simple” or “adult” with “advanced”.
A Primary student can understand sophisticated causal relationships when vocabulary is accessible. An adult can be a complete beginner in algebra or coding.
Ask for explanation based on prerequisites and output, not demographic stereotype.
The Simplification Fidelity Test
After SI gives a simple explanation, ask:
- Which parts are literally true?
- Which parts are analogy or approximation?
- Where does the simplification stop working?
- What formal concept will replace this shortcut later?
This turns a simple explanation into a bridge instead of a misconception.
The Analogy Contract
An analogy should state the mapping.
“This part of the analogy corresponds to X. This other part corresponds to Y. The analogy stops working when Z.”
If SI cannot explain the mapping, the analogy may be decorative rather than instructive.
Use analogies to enter a concept, then return to the real terms.
Worked Example: Fractions at Five Explanation Levels
Concrete
A fraction describes equal parts of one whole. If a pizza is cut into four equal pieces and you have one, you have one quarter.
Intuitive
The denominator tells how many equal parts the whole is divided into. The numerator tells how many of those parts are being counted.
Operational
To compare fractions with the same denominator, compare numerators. With different denominators, use a common representation or equivalent fractions.
Formal
A fraction a/b represents the quotient a divided by b for b not equal to zero, and rational numbers can be expressed as ratios of integers.
Transfer and limits
Fractions can represent parts, ratios, rates and operators. Pizza imagery becomes less useful for negative fractions or algebraic rational expressions.
The explanation evolves without contradicting itself.
Worked Example: Algebraic Expansion
Concrete
Three groups of x + 2 means x + 2 repeated three times.
Intuitive
The multiplier outside a bracket applies to every term inside.
Operational
Multiply the outside factor by each term, preserve signs, then combine like terms.
Formal
The distributive property states a(b + c) = ab + ac.
Transfer
Apply the same property with negative factors, variables and nested expressions.
Worked Example: Photosynthesis
Concrete
Plants use light energy to help build sugar from carbon dioxide and water.
Intuitive
Light supplies energy; carbon dioxide and water supply matter used to form glucose, with oxygen released.
Operational
Identify inputs, location, energy source and products. Connect the process to gas exchange and plant growth.
Formal
Use the balanced chemical equation and explain energy conversion and biochemical stages at the required level.
Transfer
Explain plant mass increase, food-chain energy entry and relationships with respiration.
Worked Example: Coding Functions
Concrete
A function is a named piece of instructions you can use again.
Intuitive
It can receive information, perform work and give a result back.
Operational
Define parameters, perform logic, return a value, then call the function where needed.
Formal
Discuss scope, parameters, return values, pure versus side-effecting behaviour and type expectations where relevant.
Transfer
Refactor repeated code into functions and decide which information belongs in parameters versus shared state.
Worked Example: Opportunity Cost
Concrete
If you spend your only free hour on one activity, you cannot spend the same hour on another.
Intuitive
The real cost of a choice includes the best alternative you give up.
Operational
For each option, identify the scarce resource and the best displaced use.
Formal
Explain opportunity cost in economic decision-making and distinguish it from accounting cost or sunk cost.
Transfer
Apply to money, time, attention, course choices and project portfolios.
The Explanation Compression Test
Ask SI to explain the idea in one sentence, one paragraph and one page.
Compare what survives.
The one-sentence version should preserve the core relationship. The paragraph should add mechanism. The page should add boundaries, examples and exceptions.
If the one-sentence version becomes false, the compression is too aggressive.
The Expansion Test
Start with the simple explanation and ask what new precision each advanced level adds.
This helps the learner see why technical language exists rather than experiencing it as arbitrary complexity.
A strong explanation can show the path from intuition to formalism.
The Definition–Mechanism–Example Triangle
A robust explanation usually contains three elements.
Definition: what the thing is.
Mechanism: how or why it works.
Example: what it looks like in a case.
If one is missing, the learner may know words without mechanism or examples without generality.
Explain to Different Audiences Without Stereotyping Them
Instead of “explain to a ten-year-old”, use functional descriptions.
“Explain to a learner who knows multiplication but has never seen algebra.”
“Explain to a manager who needs decision implications but not implementation detail.”
“Explain to a programmer who knows Python but is new to databases.”
“Explain to a parent who needs the learning mechanism and home support, not technical pedagogy.”
This gives SI a knowledge boundary instead of a stereotype.
The Audience Knowledge Card
Already knows: __.
Does not yet know: __.
Needs to do: __.
Technical vocabulary allowed: __.
Maximum depth: __.
Source frame: __.
A short card produces better explanation calibration than vague labels such as beginner or expert.
The Explain-Then-Test Rule
Every explanation should produce a learner action.
After the explanation, ask for one of:
- teach-back;
- prediction;
- worked example;
- classification;
- changed application;
- counterexample;
- boundary case;
- comparison with a neighbouring concept.
If the learner only says “I understand”, the tutor has weak evidence.
The One-Example Trap
An example can explain and mislead at the same time.
Learners may remember surface features and mistake them for the rule.
Use at least one contrasting example or boundary case after the first explanation.
SI makes this inexpensive.
The Explanation Error Audit
When an explanation feels clear but later performance fails, inspect the explanation.
Did it omit a prerequisite?
Did the analogy overreach?
Did it give procedure without mechanism?
Did it use advanced vocabulary too early?
Did it hide an exception?
Did the learner recognise but not retrieve?
Repair the explanation design, not only the learner.
The False Simplicity Warning
Some ideas are genuinely difficult.
A clear explanation can reduce unnecessary complexity but cannot eliminate all abstraction, prerequisite knowledge or practice.
If SI produces a beautifully simple rule that seems to solve a complex field in one sentence, ask what was omitted.
Simple is good. Simplistic is not.
A Copyable Multi-Level Explanation Prompt
“Explain this concept to me in levels without changing the underlying truth. First ask what I already know and what I need to do with the concept. Then give: 1) one concrete example, 2) an intuitive explanation, 3) the operational rule or mechanism, 4) the formal definition or model at the appropriate depth, and 5) one transfer example plus one limitation or boundary. Label any analogy and say where it stops working. After each level, ask me to explain or apply before moving higher.”
The Multi-Level Explanation Checklist
- Does the explanation match prerequisites?
- Is technical vocabulary introduced only when useful?
- Does simplification preserve the mechanism?
- Are analogies mapped and bounded?
- Is there a formal version when needed?
- Does the learner explain back?
- Is there a changed application?
- Are boundary cases visible?
- Is the source frame correct?
- Does each higher level add precision rather than contradiction?
Explanation Level Is Not the Same as Difficulty
A beginner explanation is not automatically easy, and an advanced explanation is not automatically difficult.
A beginner may need a demanding idea explained with familiar vocabulary. An expert may need a concise but technically dense explanation because the prerequisites are already available.
Level describes what the explanation assumes and what it adds—not how intelligent the learner is.
Ask SI to state the prerequisites assumed by each level. This makes it possible to choose the right version without stereotyping the learner.
The Prerequisite Vector
Instead of one label such as beginner, describe prerequisites across several dimensions.
Vocabulary: which technical terms are already familiar?
Concepts: which underlying ideas are stable?
Notation: which symbols or representations can the learner read?
Procedure: which methods can the learner execute?
Abstraction: how comfortable is the learner reasoning beyond concrete examples?
Output: what must the learner be able to do after the explanation?
This vector creates a much better explanation target than age or generic ability labels.
The Output-First Explanation Rule
Ask what the learner must do with the idea.
If the goal is to recognise a term, a short definition may be enough.
If the goal is to solve problems, the explanation needs operational structure.
If the goal is to critique research, the explanation needs assumptions, evidence and limitations.
If the goal is to build a system, the learner needs interfaces, failure modes and implementation detail.
The required output determines the required depth.
The Abstraction Ladder
Many difficult ideas become accessible when SI moves deliberately between concrete and abstract forms.
Object: one physical or familiar example.
Pattern: what stays the same across several examples.
Relationship: how the parts affect one another.
Rule: the general statement.
Representation: diagram, equation, code, model or formal notation.
System: how the rule interacts with neighbouring ideas.
Ask SI to move one rung at a time.
If the learner cannot describe the pattern across two examples, formal notation may be premature. If the learner can operate only with one concrete object, transfer has not yet occurred.
The Zoom Rule
An explanation can zoom in or out.
Zoom in when the learner needs mechanism: what exactly changes at this step?
Zoom out when the learner loses purpose: what role does this step play in the larger system?
SI can switch zoom instantly, which is one of its strongest explanatory capabilities.
Prompt: “Explain this one zoom level closer to the mechanism, then one zoom level wider to the system. Show what new information appears at each zoom.”
What, Why, How and When
A complete explanation can answer four different questions.
What?
Definition or identity.
Why?
Purpose, cause or mechanism.
How?
Procedure, implementation or sequence.
When?
Conditions, boundaries and situations where the idea applies.
A learner can know what something is without knowing why it works. They can know how to execute without knowing when to choose the method.
Ask SI which of the four dimensions is missing from the learner’s current explanation.
Formalism Should Arrive with a Reason
Technical notation exists because it compresses relationships, removes ambiguity or enables operations that ordinary language handles poorly.
When introducing a formula, symbol or formal definition, ask SI to explain why the formalism is useful.
Example: algebraic notation is not merely harder arithmetic. It represents general relationships and unknown values compactly.
Example: a function definition in programming creates reusable behaviour with explicit input and output.
Example: probability notation helps distinguish events, conditions and distributions precisely.
Formalism feels less arbitrary when the learner sees the problem it solves.
The Formalism Translation Test
Move in both directions.
Can the learner translate ordinary language into notation?
Can the learner translate the notation back into ordinary language?
If one direction fails, the learner may be manipulating symbols without meaning or understanding meaning without command of the required representation.
Worked Example: Probability
Concrete
A bag has ten counters and three are red. If one counter is chosen at random, three of the ten equally possible counters produce a red result.
Intuitive
Probability describes how likely an event is under a defined model of possible outcomes.
Operational
For equally likely outcomes, favourable outcomes divided by total outcomes can be used. For more complex situations, account for conditional or unequal probabilities.
Formal
Introduce events, sample spaces, probability measures and conditional probability at the appropriate mathematical level.
Boundary
“Favourable over total” is not a universal method when outcomes are not equally likely. The beginner shortcut needs a stated boundary.
Worked Example: Gravity
Concrete: dropped objects move toward Earth.
Intuitive: masses attract one another; near Earth’s surface that attraction produces a downward acceleration.
Operational: use gravitational field strength or acceleration in the school-level model.
Formal: introduce Newtonian gravitation and, at a much higher level, discuss where general relativity changes the model.
Boundary: “gravity is a force pulling things down” can be useful early but becomes incomplete outside the near-Earth context.
Worked Example: Recursion
Concrete: solving a smaller version of the same problem until reaching a simple stopping case.
Intuitive: a recursive function calls itself on a problem that is closer to the base case.
Operational: define a base case and a recursive step that reduces the problem.
Formal: reason about recursive definitions, stack behaviour, termination and complexity.
Transfer: compare recursion with iteration and decide when one representation is clearer.
Worked Example: Inflation
Concrete: the same amount of money buys fewer typical goods after prices rise broadly.
Intuitive: inflation describes an increase in the general price level rather than one product becoming more expensive.
Operational: compare a price index over time and distinguish nominal from real change.
Formal: discuss index construction, weighting, measurement limitations and different inflation measures at the appropriate economics level.
Boundary: one price increase is not enough to establish broad inflation.
Worked Example: Thesis Statements
Concrete: the thesis is the main answer your essay will defend.
Intuitive: it gives the reader the controlling claim rather than only naming the topic.
Operational: make a specific, arguable claim that the body paragraphs can support.
Formal: distinguish thesis, research question, claim scope, qualification and argumentative burden.
Transfer: rewrite a weak topic statement into a claim across several subjects.
The Analogy Failure Audit
Analogies fail in predictable ways.
Surface capture: learner remembers the story but not the concept.
Overextension: learner applies analogy features that do not map to reality.
False causality: analogy suggests a mechanism that is not actually present.
Missing boundary: analogy works only in a narrow case but is presented as general.
Ask SI to list the mapping and the non-mapping explicitly.
Contrast Beats Analogy When Concepts Are Neighbours
If two concepts are easily confused, contrast them directly.
Mass versus weight.
Correlation versus causation.
Function return versus print.
Theme versus thesis.
Mean versus median.
SI can create paired examples where only the decisive feature changes.
The Explanation Dependency Map
A difficult concept may need several prerequisite explanations in sequence.
Ask SI to map them but explain only the missing ones.
For calculus, limits may precede derivatives. For genetics, cell division and DNA may precede inheritance mechanisms. For databases, tables and keys may precede joins. For statistics, distributions and variability may precede inference.
A dependency map prevents the learner from receiving an advanced explanation built on invisible gaps.
The Explanation Prerequisite Gate
Before moving higher, ask one short question that demonstrates the prerequisite.
If the prerequisite fails, step back without treating that as failure. The map has simply found the correct entry point.
Explain the Same Idea to Different Professional Roles
Different professional audiences need different detail because their decisions differ.
A manager may need impact, trade-offs and risks.
An engineer may need architecture, interfaces and failure modes.
A finance colleague may need cost, uncertainty and assumptions.
A customer may need capability, limitation and action.
The core facts should remain consistent while emphasis changes.
The Role Translation Prompt
“Explain the same mechanism to three roles. Keep the factual core identical. Change only the vocabulary, detail and implications each role needs for its decision. Then show what information was omitted from each version and why.”
This is useful for professional communication as well as learning.
Explanation by Question Type
Sometimes the learner does not need another level. They need a different question answered.
Definition: what is it?
Mechanism: why does it happen?
Procedure: how do I do it?
Selection: when should I use it?
Boundary: when does it fail?
Comparison: how is it different from the nearby concept?
Transfer: where else does it apply?
Ask the learner which question remains unanswered before increasing explanation length.
The Explanation State Machine
An idea can move through states: unfamiliar → recognised → intuitive → operational → formal → transferable → integrated.
The explanation should match the current state.
If the learner is unfamiliar, give an entry point. If intuitive but not operational, provide method. If operational but not formal, add definition and boundary. If formal but brittle, add transfer and integration.
Do not start over from Level 1 every time the learner returns.
Explanation Drift
Over repeated conversations, an explanation can drift away from the source or the original mechanism.
Keep a stable formal reference for important concepts and regenerate simpler versions from that reference rather than from older paraphrases.
This reduces the telephone-game effect of repeatedly simplifying a simplification.
Source-Grounded Explanation
When a definition or method must match a syllabus, standard or documentation, supply that source.
Ask SI to separate source-defined terms from its own teaching language.
Example: “Use the official definition exactly, then explain it in plain language, then show an example. Do not change the formal meaning.”
This approach is especially useful for law, science, professional standards, software and assessment rubrics.
The Explanation Provenance Rule
If an important explanation introduces a factual claim beyond the supplied source, label it and verify it where needed.
A teaching explanation should not silently expand into unsourced general knowledge simply because the prose is smooth.
Worked Example: Derivatives
Concrete: measure how quickly a quantity is changing around one point.
Intuitive: the derivative is the instantaneous rate of change and can be visualised as the slope of the tangent line.
Operational: use differentiation rules to calculate derivatives of familiar functions.
Formal: define derivative through a limit of difference quotients.
Transfer: connect derivatives to velocity, optimisation, marginal change and function behaviour.
Boundary: the simple “slope” intuition must be refined for functions that are not differentiable at a point.
Worked Example: Correlation and Causation
Concrete: two things can move together without one causing the other.
Intuitive: correlation describes association; causation describes a mechanism where changing one factor contributes to changing another.
Operational: ask about confounding variables, temporal order, alternative explanations and study design.
Formal: introduce statistical association, causal inference assumptions and experimental or quasi-experimental designs at the appropriate level.
Transfer: critique a headline that turns a correlation into a causal claim.
Worked Example: Neural Networks
Concrete: a system adjusts many numerical connections so inputs produce more useful outputs.
Intuitive: layers transform representations, and training changes parameters to reduce error on examples.
Operational: inputs pass forward, loss measures mismatch and optimisation updates parameters.
Formal: introduce weighted sums, nonlinear activations, loss functions, gradients and backpropagation at the appropriate mathematical level.
Boundary: the “brain-like neurons” analogy is limited and should not be treated as a literal model of biological cognition.
The Explain-Back Rubric
When the learner explains the idea back, check five elements.
- Correct definition or central relationship.
- Mechanism or reason.
- One valid example.
- One condition or boundary.
- No contradiction between intuitive and formal versions.
SI can score or critique the rubric, but the learner should produce the explanation unaided first.
The Explanation Transfer Test
Understanding survives when the learner can use the idea outside the example that taught it.
Ask for a new case, a counterexample, a comparison or an application.
If the learner can only repeat the explanation, they may have learned the wording rather than the concept.
SI should always connect explanation to transfer when the skill matters beyond recognition.
The Explanation Selection Rule
Do not ask SI for all explanation levels unless you need them.
Select the lowest level that enables the required output and add higher precision only when the learner is ready or the task demands it.
This makes explanations shorter and reduces cognitive load.
More levels are a menu, not a requirement.
The Explanation Stop Rule
Stop explaining when the learner can perform the next productive task.
Do not continue adding examples because the conversation is enjoyable.
Once the learner can retrieve and apply, switch from explanation to practice.
The purpose of an explanation is to make independent thinking possible, not to become the main activity forever.
Progressive Formalisation
Difficult ideas often become learnable when the tutor moves through three passes: meaning, structure and formalism.
Meaning: what is happening in ordinary language?
Structure: what are the parts and relationships?
Formalism: how is the idea expressed precisely using notation, definitions or technical language?
For algebra, meaning may begin with balancing unknown quantities, structure with operations preserving equality, and formalism with symbolic equations.
For programming, meaning may begin with reusable behaviour, structure with input–process–output, and formalism with function syntax, parameters and return values.
For economics, meaning may begin with limited resources and choices, structure with trade-offs, and formalism with opportunity cost, marginal analysis and models.
SI can change levels instantly. The learner should move only when the previous representation is stable enough to support the next one.
Translate the Same Idea Across Representations
One of the strongest tests of understanding is translation.
- words → diagram;
- diagram → equation;
- equation → words;
- code → plain-language steps;
- story → abstract principle;
- principle → new example.
If the learner can translate among representations, understanding is usually stronger than if they can only repeat one format.
The Representation Gap
Sometimes the learner understands the idea in one representation but fails when the representation changes.
A student may understand a ratio diagram but not an algebraic expression. A coder may understand the algorithm verbally but not implement it. A science learner may recognise the word equation but fail to explain the particle mechanism.
Ask SI to translate the same concept across two representations and make the learner perform the reverse translation.
Analogy: Useful Only When Its Boundaries Are Visible
Analogies compress unfamiliar structure into familiar structure. They can accelerate understanding and also create durable misconceptions.
Ask for three parts: what the analogy captures, what it omits and where it becomes misleading.
For electric current as water flow, the analogy may help with flow and resistance but can mislead if treated as a literal description of charge behaviour.
For a function as a machine, the analogy helps with input and output but may hide scope, side effects, higher-order functions and composition.
For memory as storage, the analogy may help with retrieval but can hide reconstruction and interference.
The tutor should eventually remove the analogy and restate the real mechanism.
The Analogy Exit Test
Ask the learner to explain the concept without the analogy. If they cannot, the analogy has become a substitute for the concept rather than a bridge to it.
Explain by Contrast
Two easily confused ideas often become clearer when explained together.
Mass versus weight. Correlation versus causation. Revenue versus profit. Parameter versus argument. Reluctant versus resistant. Accuracy versus precision.
Ask SI for a contrast table with definition, shared features, decisive difference, common mistake and one classification example.
Then remove the table and test changed examples.
Explain by Counterexample
A counterexample exposes the boundary of an oversimplified rule.
If the learner believes “multiplication always makes numbers larger”, use multiplication by a fraction. If they believe “every function that produces output must print”, use a function that returns a value silently. If they believe “more study hours always improve results”, compare unfocused rereading with targeted retrieval.
Ask SI for the smallest counterexample capable of breaking the learner’s current misconception.
Explain by Causal Chain
For processes, a causal chain is often clearer than a descriptive paragraph.
What initiates the process? What changes next? What mechanism connects the steps? What is the output? What happens if one step is removed?
This method is useful for science, history, economics, biological systems and operational workflows.
Worked Example: Programming Functions at Six Levels
Level 1: a function is a named block of instructions you can reuse.
Level 2: it can take information in, process it and give a result back.
Level 3: identify parameters, function body, call and return value.
Level 4: introduce scope, side effects, type signatures and composability as appropriate.
Level 5: contrast returning a value with printing it, and show why confusing the two breaks composition.
Level 6: decompose a larger program into functions and reason about the interfaces among them.
The invariant is reusable transformation. Each level adds precision and system context.
Worked Example: Photosynthesis at Six Levels
Level 1: plants use light energy to make food from carbon dioxide and water.
Level 2: light energy enables the plant to build glucose, connecting inputs and outputs.
Level 3: use the word equation and explain where each input comes from.
Level 4: introduce the chemical equation, chlorophyll and cellular location at the appropriate syllabus level.
Level 5: correct the misconception that plants get their food mass directly from soil.
Level 6: connect photosynthesis with respiration, biomass, food chains and gas exchange.
Worked Example: Opportunity Cost at Six Levels
Level 1: choosing one use for a limited resource means giving up another use.
Level 2: spending Saturday studying may mean giving up sport or rest.
Level 3: opportunity cost is the best forgone alternative, not every alternative.
Level 4: connect it to scarcity, marginal choice and resource allocation.
Level 5: contrast it with sunk cost.
Level 6: apply it to career, education, capital allocation and family scheduling.
Worked Example: A Historical Explanation
Suppose the learner asks why empires decline.
At a simple level, explain that governing, funding and defending a large system can become difficult.
At the next level, introduce interacting pressures: revenue, military commitments, political legitimacy and administrative capacity.
At a deeper level, distinguish structural pressures, triggers, historical evidence and competing interpretations.
At the highest learning level, compare two empires where similar pressures produced different outcomes and ask what variable changed the trajectory.
The explanation becomes deeper by adding causal structure and evidence limits, not merely more dates.
The Oversimplification Audit
After a simple explanation, ask four questions.
- What important detail did we remove?
- Which condition makes the simple rule true?
- Where would the rule fail?
- What misconception could this wording create?
A good simplified explanation is a stepping stone toward a more accurate model, not a replacement for one.
The Complexity Audit
Advanced explanations can fail by including detail that does not change the learner’s next task.
Ask SI to label each technical detail: essential now, useful later or enrichment only.
Remove details that consume working memory without improving the learner’s model.
Depth should follow purpose, not the amount of knowledge available to the model.
Age, Expertise and Explanation Level
Age is not the same as knowledge level.
A young learner can understand a sophisticated idea when prerequisites are strong. An adult can be a true beginner in a new domain.
For children, use shorter steps, familiar examples and age-appropriate language without speaking down to them.
For adults, connect new ideas to existing mental models and real-world use.
For experts, reduce introductory explanation and increase precision, assumptions, edge cases and trade-offs.
Ask SI to calibrate to prerequisite knowledge rather than demographic stereotype.
Beginner, Intermediate and Expert Comparison
Ask SI to explain the same concept three times: beginner, intermediate and expert.
Then compare what stayed invariant and what changed.
The invariant reveals the concept core. The changing detail reveals the scaffolding each level needs.
The Explanation Diagnostic
When the learner says “I still don’t get it”, do not automatically simplify again.
Check whether the problem is an unknown word, missing prerequisite, mismatch between example and rule, lack of production, weak method recognition or an existing misconception.
The next explanation should target the failure type.
Repeated simplification without diagnosis can make the explanation less accurate while leaving the real gap untouched.
The Explain-Back Test
Ask the learner to explain the idea without reusing SI’s wording.
Then ask one “why”, one “when” and one “what if” question.
Why does it work? When would you use it? What changes if one condition changes?
These questions test mechanism, recognition and boundary.
The Zoom-In / Zoom-Out Method
A learner may understand the broad idea but lose one mechanism.
Ask SI to zoom in only on that step, explain it in detail, then zoom out and reconnect it to the whole system.
This prevents local explanation from becoming disconnected fragments.
Prompt: “I understand the overall process but not __. Explain only that mechanism, then show exactly where it sits in the larger sequence.”
The Compression / Expansion Method
Ask SI first to compress the concept into one sentence. Then expand each important word in that sentence.
This creates a hierarchy of detail rather than one flat wall of text.
It is especially useful for complex topics where the learner needs orientation before depth.
How to Verify a Difficult Explanation
For important technical material, compare the explanation with a trusted source.
School learning: syllabus, textbook and teacher material. Coding: current documentation. Science: authoritative educational references or primary material at the appropriate level. Professional subjects: standards, current documentation and expert guidance.
Ask SI which parts are formal definitions and which parts are teaching interpretations.
A learner should become capable of checking explanations, not only consuming them.
The Explanation Source Card
Concept: what is being explained.
Level: intended depth.
Source: what defines correctness.
Simplification: what detail was omitted.
Boundary: where the explanation fails.
Transfer task: what proves understanding.
This card is useful for concepts that recur across a course or project.
The Explanation-to-Learning Handoff
An explanation is complete only when control returns to the learner.
- teach it back;
- solve a changed problem;
- classify an example;
- create an example;
- identify an error;
- compare with a nearby concept;
- apply it in a new context.
Without this handoff, the interaction remains consumption.
The Explanation Dependency Test
If the learner can explain the concept only by repeating SI’s phrases, support dependence may remain high.
Close the explanation, wait, then ask the learner to reconstruct the concept in their own representation.
Independent reconstruction is stronger evidence than immediate paraphrase.
A 14-Day Multi-Level Explanation Practice
Day 1: explain one concept at Level 1 and Level 4.
Day 2: identify the invariant across levels.
Day 3: create one analogy and its failure points.
Day 4: translate words into a diagram or formula.
Day 5: explain by contrast.
Day 6: explain with a counterexample.
Day 7: run an explain-back test.
Day 8: diagnose a failed explanation.
Day 9: zoom into one difficult mechanism.
Day 10: remove unnecessary technical detail.
Day 11: add formal precision to an oversimplified explanation.
Day 12: verify an explanation against a trusted source.
Day 13: create a transfer task.
Day 14: save the explanation prompt that best fits your learning style.
Common Explanation Failure Modes
Simpler but wrong: repair by stating mechanism and boundary.
Technical but empty: add a concrete example and map the terms.
Analogy becomes the concept: exit the analogy and restate formally.
Too many examples: ask what invariant the examples share.
Too much detail too early: separate essential now from useful later.
Learner only recognises: require production and transfer.
Explanation ignores the misconception: diagnose before simplifying again.
No source check: verify important technical claims.
The Explanation Quality Checklist
- Does the explanation match the learner’s starting knowledge?
- Is the mechanism preserved?
- Are technical terms introduced when needed?
- Does the analogy state its limits?
- Can the learner translate to another representation?
- Are edge cases visible at the right stage?
- Can the learner explain it back?
- Can the learner use it on a changed task?
- Is the explanation source-aligned?
- Is support removed after understanding improves?
The Final Explanation Rule
Control vocabulary, abstraction and mechanism depth separately.
Make the explanation concrete enough to enter, accurate enough to trust, formal enough to transfer and bounded enough to avoid misconception.
The goal is not to make a difficult idea sound easy. It is to make the difficult idea progressively thinkable without changing what the idea really is.
Explanation Diagnostics: When “I Don’t Understand” Is Too Broad
Ask what part of the explanation failed.
Vocabulary failure: a key term is unknown.
Reference failure: the explanation refers to a concept the learner has never met.
Relationship failure: individual terms are known but their connection is unclear.
Mechanism failure: the learner knows what happens but not why.
Procedure failure: the learner understands the idea but cannot execute.
Selection failure: the learner cannot recognise when to use it.
Boundary failure: the learner overgeneralises the rule.
Transfer failure: the learner cannot apply the idea in a changed context.
SI can use this diagnostic to choose the next explanation level instead of repeating the same explanation with different words.
The “Show Me Where You Lost Me” Rule
Ask the learner to identify the last sentence or idea that still made sense.
The tutor should restart from that boundary, not from the beginning of the topic.
This is especially useful in technical subjects where one missing relationship can make every later sentence feel opaque.
It also prevents explanation fatigue from hearing the parts already understood again.
The Explanation Branches
There is rarely one best explanation. Build branches for different failure types.
If vocabulary is missing: define terms and use one example.
If mechanism is missing: show the causal or logical sequence.
If procedure is missing: model the steps and explain why each step exists.
If method selection is missing: use contrast cases rather than another worked solution.
If boundary is missing: provide counterexamples and failure cases.
If transfer is missing: remove surface cues and change context.
The branch system makes explanations adaptive without making them longer by default.
The Explanation Decision Tree
Ask one question at a time.
Can the learner define the concept? If no, begin with vocabulary and concrete example.
Can the learner explain the relationship? If no, use mechanism.
Can the learner perform the procedure? If no, use operational explanation.
Can the learner recognise when to use it? If no, use contrasts.
Can the learner explain limits? If no, use boundary cases.
Can the learner transfer? If no, change the surface and context.
This tree turns explanation into diagnosis rather than a monologue.
Novice Language and Expert Language
Experts compress knowledge into terms that carry many relationships at once.
A novice hears the term but not the compressed structure behind it.
“Opportunity cost”, “derivative”, “recursion”, “homeostasis”, “normalisation”, “causal inference” and “thesis” all compress substantial conceptual structure.
Ask SI to unpack the expert term into the relationships it contains, then gradually reintroduce the term after the relationships are understood.
The Expert Compression Test
For an advanced learner, run the process in reverse.
Give the detailed mechanism and ask the learner to compress it into the expert term or formal representation.
This checks whether the learner understands why the shorthand is legitimate.
Explanation and Misconception Repair
If a learner holds a wrong model, a clean explanation may simply be absorbed into the wrong model.
Expose the misconception first.
Example: learner believes heavier objects always fall faster.
Do not merely explain gravity. Ask the learner to predict two objects under controlled conditions and discuss what variables actually matter.
Example: learner believes a larger denominator means a larger fraction.
Use one whole divided into more equal parts and ask which part becomes smaller.
Explanation becomes stronger when it replaces a model rather than layering words on top of it.
The Misconception Contrast
Ask SI to generate one example where the misconception and correct model predict the same result and one where they differ.
The differing case is diagnostic.
This technique is efficient because it targets the boundary between the wrong and correct models.
Explanation Through Multiple Representations
One idea can be represented in words, symbols, diagrams, tables, code, stories or physical examples.
Ask SI to show the same relationship in two or three forms.
For proportionality: words, table and equation.
For a chemical reaction: word equation, symbolic equation and particle description.
For a function: verbal rule, input-output table and code.
The learner should translate between representations.
Representation Mismatch
Sometimes the concept is understood in one representation but not another.
A student understands a graph visually but cannot express the relationship algebraically.
A programmer understands pseudocode but cannot implement syntax.
A language learner recognises written vocabulary but cannot produce it orally.
The explanation should target the translation, not reteach the whole concept.
The Executive Explanation
An executive explanation is not a beginner explanation.
It assumes the reader may understand complex ideas but needs only the mechanism, implication, uncertainty and decision relevance.
Ask SI for four fields: what it is, why it matters, what could go wrong, and what decision it affects.
This is useful for managers, parents, clients and decision-makers who do not need implementation detail.
The Technical Explanation
A technical explanation should expose assumptions, interfaces, formal definitions, edge cases and failure modes.
Do not make it verbose simply to sound technical.
Technical depth should correspond to what the reader needs to implement, verify or critique.
The Teaching Explanation
A teaching explanation includes sequencing and learner checks.
It should anticipate misconception, choose examples, ask for retrieval and test transfer.
That makes it different from a reference explanation, which may be accurate but not pedagogically sequenced.
The Explanation Fidelity Audit
Compare the simple, operational and formal versions side by side.
Do they preserve the same central relationship?
Does the simple version introduce a false causal story?
Does the formal version add conditions rather than contradict the intuition?
Does the example illustrate the rule or accidentally create a narrower rule?
Ask SI to identify any sentence in the simpler version that would need correction later.
A good simple explanation can be expanded. A bad one must be unlearned.
The “No Future Unlearning” Rule
A beginner explanation can omit detail, but it should avoid teaching a claim that future learning must reverse.
“Electrons orbit the nucleus like planets” may be useful as historical imagery but can become misleading if presented as literal modern physics.
“The brain is a computer” can illuminate information processing but should not be treated as a complete biological model.
Ask SI to mark approximations explicitly so the learner knows which parts will later be refined.
The Explanation Sequence for a New Topic
1. Activate prior knowledge. What related idea is already stable?
2. Introduce one concrete case.
3. Extract the invariant relationship.
4. Name the concept.
5. Introduce operational or formal representation.
6. Contrast with a neighbouring concept.
7. Test boundary.
8. Require transfer.
This sequence is not mandatory for every topic, but it prevents definitions from arriving before meaning.
The Explanation Sequence for an Existing Misunderstanding
1. Ask learner’s current explanation.
2. Identify the wrong relationship.
3. Create a contrast case where the misconception fails.
4. Explain the corrected mechanism.
5. Learner explains it back.
6. Test transfer.
This is more efficient than simply presenting the correct explanation on top of the wrong model.
A Full Case Study: Explaining Simultaneous Equations
Learner state: can solve single linear equations but has never seen two equations describing the same unknowns.
Concrete explanation: two clues describe the same hidden quantities.
Intuitive explanation: both equations must be true at the same time, so the solution is the pair that satisfies both.
Operational explanation: substitution replaces one unknown using an equivalent expression; elimination combines equations to remove one unknown.
Formal explanation: solution is the intersection of two linear equations or the ordered pair satisfying the system.
Contrast: one pair of equations is easiest by elimination; another by substitution.
Transfer: word problem where the learner constructs both equations and chooses the method.
The learner sees method as a consequence of structure rather than a memorised chapter procedure.
A Full Case Study: Explaining Supply and Demand
Concrete: when many people want a limited number of tickets, buyers may be willing to pay more.
Intuitive: demand represents willingness to buy at different prices; supply represents willingness to sell.
Operational: shifts in demand or supply change equilibrium price and quantity under the simplified model.
Formal: introduce curves, equilibrium, elasticity and ceteris paribus assumptions.
Boundary: real markets may contain price controls, strategic behaviour, imperfect information and other complications.
Transfer: analyse a new scenario and identify whether supply, demand or both shifted.
A Full Case Study: Explaining Database Joins
Concrete: combine information from two lists using a shared identifier.
Intuitive: a join matches rows according to a relationship between keys.
Operational: choose inner, left, right or other joins depending on which unmatched rows should remain.
Formal: discuss relational algebra, cardinality, key constraints and null behaviour as required.
Boundary: “joining is like merging spreadsheets” can hide duplicate-key and cardinality problems.
Transfer: predict row counts before executing several joins.
The Explanation Memory Test
After a delay, ask the learner to reconstruct the explanation.
Can they state the central relationship? Can they produce an example? Can they identify one boundary?
If they remember the analogy but not the mechanism, redesign the explanation.
If they remember the formal definition but cannot apply it, add operational and transfer practice.
The Explanation Durability Rule
A durable explanation gives the learner a structure they can reconstruct, not a paragraph they must memorise.
Good structures include causal sequences, comparison tables, invariant relationships, diagrams and formal definitions connected to examples.
SI should help the learner compress the explanation into a reusable mental structure.
The Explanation Review Card
Concept: what is being explained.
Learner prerequisites: what is assumed.
Required output: what the learner must do.
Entry representation: concrete, intuitive, symbolic or other.
Formal core: precise definition or mechanism.
Analogy: if used, mapping and limit.
Boundary: where the simple rule stops.
Transfer test: changed application.
Source: what anchors the formal meaning.
This card lets the same concept be explained differently to different learners without losing fidelity.
The Explanation Independence Test
Remove SI and ask the learner to explain the idea to someone else.
Can they choose an example, state the mechanism and answer a reasonable “why?” question?
Teaching another person is not proof of complete mastery, but it reveals whether the learner owns the explanation rather than recognising the model’s wording.
The Final Explanation Rule
A good explanation at a lower level should be something you can later add precision to, not something you must erase.
A good explanation at a higher level should reveal why the earlier version worked and where it was incomplete.
The levels should form one staircase of understanding, not five unrelated stories about the same topic.
The Misconception-First Explanation
Sometimes the learner does not lack an explanation. They already have an explanation that is wrong.
In that case, adding a cleaner explanation may not be enough because the old model keeps interpreting the new information.
Ask the learner to predict what should happen before the tutor explains. Then compare the prediction with the actual case.
Example: learner believes multiplying always makes a number bigger. Use multiplication by a fraction. Learner believes heavier objects fall faster in ordinary conditions because “more weight pulls harder”. Compare carefully controlled cases. Learner believes every function call must print something. Use a returned value that is not printed.
The explanation should directly replace the broken mechanism, not simply place correct words beside it.
The Prediction–Explanation–Retest Loop
Predict: learner states what they expect.
Explain: tutor addresses the mismatch.
Retest: learner predicts a changed case.
Generalise: learner states the improved rule.
This loop is especially strong for persistent misconceptions because it creates visible conflict between the old model and the new evidence.
The Cognitive-Load Check
An explanation can be correct and still be badly timed.
If the learner must hold unfamiliar notation, new vocabulary, a multi-step process and a new example in mind simultaneously, the explanation may exceed working capacity.
Break the load apart. Introduce one representation, stabilise it, then add the next layer.
Ask SI: “Which parts of this explanation are new at the same time? Reorder them so only one major new element is introduced per step.”
Chunking
Experts see chunks that beginners see as many separate details.
A fluent algebra student sees “expand brackets” as one chunk. A beginner sees signs, coefficients, variables and multiplication separately.
SI can help identify useful chunks and name them, but the learner still needs repeated retrieval and application before the chunk becomes automatic.
Good explanations create chunks that can later be combined into larger reasoning sequences.
Explain the Boundary, Not Only the Rule
Beginners often learn a rule without learning when not to use it.
Ask SI to include one positive example, one near-miss and one counterexample.
For correlation and causation: one genuine causal mechanism, one correlation with plausible confounding, one unrelated coincidence.
For a grammatical structure: one correct context, one context where another tense or form is more appropriate.
For a mathematical method: one problem where the method is efficient and one where another method is better.
Boundaries turn knowledge into discrimination rather than memorised slogans.
The Near-Miss Example
A near-miss looks similar to the target concept but differs in one decisive feature.
Near-miss examples train recognition. They answer the learner’s hidden question: “How do I know this is the concept and not something nearby?”
Ask SI to generate near-misses whenever the learner knows the definition but misclassifies real cases.
Explanation Through Multiple Examples
One example can be mistaken for the rule. Several examples reveal the invariant.
Ask for three examples that look different on the surface but depend on the same mechanism.
Then ask the learner: “What is the same across all three?”
The answer should move from surface features toward the concept core.
Example Diversity
Examples should vary along meaningful dimensions, not just change numbers or names.
For algebra, vary sign, arrangement, context and whether expansion is necessary. For vocabulary, vary tone, register and syntactic position. For coding, vary input type, side effects and composition. For economics, vary personal, business and public-policy contexts.
Diversity makes transfer visible.
The Teach-to-Someone-Else Test
Ask the learner to explain the concept to a hypothetical beginner.
The goal is not performance theatre. Teaching forces selection: which idea is essential, which example clarifies it and which details can wait?
SI can play the beginner and ask follow-up questions. The learner should answer without returning immediately to the original explanation.
If the learner cannot respond to a basic “why?” or “when would this not work?”, the model may still be shallow.
The Hostile-Beginner Test
For advanced learners, ask SI to challenge the explanation with common misconceptions, edge cases and precise questions.
The purpose is to reveal hidden gaps before the learner teaches, writes or applies the concept in a higher-stakes setting.
A Full Case Study: Explaining Probability
Level 1: probability describes how likely an event is.
Level 2: use a bag containing coloured counters and compare favourable outcomes with all possible outcomes.
Level 3: calculate simple probabilities and distinguish impossible, unlikely, even chance, likely and certain.
Level 4: use formal notation and discuss sample spaces.
Level 5: introduce dependent events, conditional probability or misleading intuition where appropriate.
Level 6: apply probability to decision-making, risk and statistical inference.
Misconception check: “A coin has landed heads five times, so tails is now more likely.” Use the explanation level appropriate to challenge the gambler’s fallacy without overwhelming the learner.
A Full Case Study: Explaining Correlation and Causation
Level 1: two things can change together without one causing the other.
Level 2: use a simple example where a third variable influences both.
Level 3: identify temporal order, mechanism and alternative explanations.
Level 4: introduce confounding, experimental design and causal inference vocabulary.
Level 5: show cases where randomised experiments are impossible and evidence must be triangulated.
Level 6: evaluate a real claim and distinguish what the evidence establishes from what remains interpretive.
A Full Case Study: Explaining Functions in Mathematics
Level 1: each allowed input is assigned an output.
Level 2: use a table or machine analogy.
Level 3: connect input–output relationships to rules and graphs.
Level 4: introduce domain, range and notation.
Level 5: show relations that are not functions and discuss one-input-one-output boundaries.
Level 6: connect functions to composition, inverse relationships and modelling.
The Explanation Version Rule
Do not keep revising one giant explanation forever.
Version the explanation for purpose: beginner orientation, operational use, formal study, transfer review.
Each version should state what background it assumes and what the learner should be able to do after reading it.
This makes explanation reusable and prevents a beginner version from being mistaken for the complete formal model.
The Explanation Continuity Rule
When a learner returns later, resume from demonstrated level rather than from the beginning.
If the learner can still use the formal rule but forgot one edge case, explain only the edge case. If only the analogy remains, rebuild the bridge to formal structure.
Good explanation systems preserve progress instead of repeatedly resetting the learner to beginner mode.
The Final Explanation Transfer Test
Take one concept and explain it at three levels without consulting the original SI answer.
Then solve or analyse a changed case. Finally, identify one boundary or counterexample.
If you can move between simple and formal representations while preserving the mechanism, the explanation has become part of your own model rather than remaining borrowed language.
That is the real goal of multi-level explanation: not easier text, but flexible understanding.
The Explanation Ownership Test
Before moving on, ask the learner to rebuild the explanation without seeing SI’s version. They should be able to choose the level themselves: one-sentence intuition, operational explanation or formal account.
If the learner can repeat wording but cannot choose which detail belongs at which level, the explanation may still be externally held.
Understanding becomes durable when the learner can reconstruct the explanation at the level the situation requires.
Frequently Asked Questions
What does “explain at different levels” mean?
It means changing vocabulary, representation, abstraction and formal depth according to the learner’s prerequisites and required output while preserving the underlying concept.
Should I ask for ELI5?
It can be a useful start, but a knowledge card is usually more precise because age does not define subject knowledge.
Can analogies cause misconceptions?
Yes. Ask where the analogy maps correctly and where it breaks.
How do I know an explanation is too simple?
If it becomes false, hides a load-bearing condition or fails when the learner sees a normal boundary case, it is too simple.
How do I know an explanation is too advanced?
If the learner is blocked by prerequisite vocabulary or concepts rather than the target idea, step down one layer.
Should every explanation include the formal definition?
Only when the learner’s goal or current stage requires it. Formalism should be available, not forced prematurely.
How do I test understanding after an explanation?
Require teach-back and a changed task without the original wording or example.
What comes next?
The next article shows how to turn an unfamiliar subject into a structured learning roadmap.
Helpful Reading
- How to Build a Personal Super Intelligence Tutor
- How to Learn Anything Faster with Super Intelligence
- How to Think More Clearly with Super Intelligence
- How to Give Super Intelligence Better Context
Explain Down, Then Build Back Up
The best explanation meets the learner where they are without pretending the destination is simpler than it is.
Start concrete when needed. Build intuition. Add operational structure. Introduce formal precision. Test transfer and boundaries.
Super Intelligence is most useful when it can move flexibly between those levels while preserving one coherent idea underneath them all.
