The Feynman Technique is one of the most well-researched methods for deep understanding. AI strengthens every step. We break down how this works, with a practical example.

Introduction

Richard Feynman — Nobel Prize-winning physicist and one of the most original thinkers of the 20th century — was known not just for his scientific achievements but for his ability to explain the most complex concepts so that anyone could understand them. His colleagues called it "the Feynman technique," though he described it more simply: if you can't explain something in plain words, you don't truly understand it.

The Feynman technique became a popular learning tool for good reason: it works. Cognitive psychology explains why — attempting to explain a topic reveals gaps in understanding more accurately than any other verification method. Reading an explanation, we think "got it." Trying to explain it ourselves, we discover we understood only part of it.

The classic implementation of the method had one limitation: you need someone to explain to, and someone who can give feedback. An imaginary child or a blank notebook can't point out a logical error or ask a clarifying question.

AI removes this limitation. For the first time, the Feynman technique can be implemented fully — with real feedback — without a tutor or a classmate.

What the Feynman Technique Is and Why It Works

The Feynman technique is based on a simple observation: there is a fundamental difference between the feeling of understanding and actual understanding. The feeling arises when we read an explanation and it seems logical. Actual understanding means we can reproduce the idea independently, explain it to another person, and apply it to a new problem.

Attempting to explain a topic switches the mode from passive recognition to active recall. In cognitive psychology, this is called "retrieval practice." It not only checks understanding but reinforces it: the act of trying to explain makes knowledge more durable.

When an explanation "breaks down" — when you can't find the words or realize you can't articulate the step from A to B — you've found a gap. This is the most valuable moment of the method: precise identification of exactly what isn't understood.

Research in educational psychology shows that students systematically overestimate their understanding of material after reading it. After attempting to explain it — the assessment becomes significantly more accurate. This effect is called the "familiarity illusion": the brain confuses recognition of text with reproduction of ideas.

The Four Steps of the Feynman Technique

Step 1: Choose a Topic and Write Down Everything You Know

Take a sheet of paper (or open a new document) and write the topic at the top. Then write everything you know about it — without looking things up, without sources, in your own words. Don't try to be structured — just unload your understanding as it is.

The goal of this step is to capture a "snapshot" of your current understanding. What you've written is what's actually in your head — not what's in the textbook.

Step 2: Explain the Topic as if to a Child

Rewrite your explanation so that someone who knows nothing about the topic could understand it — imagine explaining it to a 12-year-old.

This doesn't mean oversimplifying to the point of inaccuracy. It means removing jargon, unpacking abbreviations, using analogies, and showing concrete examples. If you can't explain a concept without specialized terminology — you probably understand it through the terms rather than through the underlying idea.

This is the step where gaps are usually discovered: "I understand A, I understand C, but I can't explain how one follows from the other" — which means B isn't understood.

Step 3: Return to the Source for the Gaps

When you've found where the explanation breaks down — return to the textbook, your notes, or another source specifically for that point. Don't re-read everything: focus on the specific gap.

After studying that point, try again to explain it in plain words. Repeat until the explanation is complete and coherent.

Step 4: Simplify and Find an Analogy

The final step is making the explanation as clear and compact as possible. Find an analogy from everyday life that captures the essence. Remove everything unnecessary.

A good test: can you explain the topic in two minutes? If not — find the point where the explanation loses its thread.

Feynman said: if you can't explain it simply, you don't understand it well enough yet. That's not a criticism — it's a diagnosis.

How AI Strengthens Each Step

The classic Feynman technique works without AI — Feynman developed it long before language models existed. But AI allows the method to be implemented more fully and provides more precise feedback at each step.

AI at Step 1 (Understanding Snapshot)

After you've written everything you know, ask AI:

Here is my understanding of the topic [topic name]:
[your text]

Don't explain the topic to me again. Instead:
1. What in my explanation is accurate?
2. Where did I understand it inaccurately or partially?
3. What did I miss entirely?
4. What is the key concept of this topic I should have mentioned?

This provides a structured assessment of your starting level — without a prior AI explanation that might "contaminate" your own understanding.

AI at Step 2 (Explanation for a Child)

After writing your simple explanation, ask AI to play a role:

Act as a smart 12-year-old I've just explained [topic] to.
Ask me questions — the kind that would genuinely arise for a child hearing this for the first time.
If something in the explanation is unclear or contradictory — point it out with a question.
Don't accept the explanation at face value: ask "but why?", "what does that mean?", "can you give me an example?"

This creates a dialogue that surfaces gaps invisible in self-review. "Why?" questions are especially valuable: they expose places where the explanation rests on an assumed premise rather than genuine understanding.

AI at Step 3 (Working Through Gaps)

When a gap surfaces in the dialogue, ask AI to help with precisely that — but through questions, not explanation:

I can't explain [specific gap].
Don't explain this to me directly. Ask guiding questions that will help me arrive at understanding on my own.
Start with the simplest possible question.

This format — Socratic dialogue — reinforces understanding far better than receiving a ready explanation. You actively build the bridge to understanding, rather than passively receiving a completed one.

AI at Step 4 (Checking the Final Explanation)

After refining your explanation to its final form, ask AI for a concluding evaluation:

Here is my final explanation of [topic] — as I would explain it to someone unfamiliar with the subject:
[your explanation]

Evaluate this explanation on three criteria:
1. Accuracy: are there any factual or conceptual errors?
2. Completeness: are there important aspects of the topic I didn't cover?
3. Clarity: are there places that would be unclear to an unprepared reader?

Don't rewrite the explanation. Only point out specific places for improvement.

A Practical Example: Topic and Method in Action

A concrete example: recursion in programming.

Step 1 — Understanding snapshot. The student writes: "Recursion is when a function calls itself. Used to traverse trees and other complex data structures."

AI snapshot assessment: "Your definition is technically correct but incomplete. You didn't mention: the base case (stopping condition), what causes infinite recursion, the call stack, or recursion vs. iteration."

Step 2 — Explanation for a 12-year-old. The student writes: "Imagine you're looking for your keys. You check one room, then go to another room and check it the same way. Recursion is when a function solves a problem by breaking it into a smaller version of the same problem, until the problem is so small it's easy to solve."

AI as a 12-year-old: "Okay, that makes sense. But how does the function know when to stop? Doesn't it just keep breaking down forever?"

This is exactly the question that surfaces the gap: the student hasn't explained the base case.

Step 3 — Working through the gap. AI asks guiding questions: "If I ask you to count down from 5 to 1 — how do you know when to stop? What condition signals 'the problem is small enough'?"

Through dialogue, the student formulates: the base case is the condition at which recursion stops — otherwise the function calls itself indefinitely.

Step 4 — Final explanation and check. The student writes a complete explanation with an analogy (Russian nesting dolls: you open one doll inside another until you reach the smallest one that can't be opened — that's the base case). AI checks it: accuracy is good; completeness — missing a mention of the call stack and overflow risks.

In one cycle of the method, the student moved from an incomplete technical definition to a working explanation with an analogy and an understanding of key risks.

Typical Mistakes When Applying the Method

Mistake 1: Reading the material before attempting to explain. If you first read a good explanation and then try to "explain it in your own words" — you're reproducing someone else's explanation, not your own understanding. The method only works if the snapshot is taken before, not after, studying the source.

Mistake 2: Treating difficulty as incompetence. If the explanation breaks down — that doesn't mean the topic is too hard or you studied poorly. It means a gap was found. That's exactly what the method is for.

Mistake 3: Asking AI to explain right away, without attempting first. If at the first difficulty you open AI for an explanation — the method doesn't work. The Feynman approach requires an attempt first, feedback only after.

Mistake 4: Skipping the analogy. The final step — finding an analogy — seems optional, but it's precisely where the depth of understanding becomes visible. If you can't find an everyday analogy, the understanding is probably still superficial.

Mistake 5: Doing it only once. The Feynman technique works better as a regular practice. One cycle will surface gaps. Repeated cycles after a few days will show whether the understanding held — or broke down again.

How OpenEd Implements the Feynman Technique

The AI Mentor in OpenEd is specifically configured for Socratic interaction mode. When a student starts on a topic, the mentor doesn't begin with an explanation — it begins with a question: "What do you already know about this? Try explaining it in your own words."

This is a direct implementation of Feynman's first step. The mentor asks clarifying questions — like the smart 12-year-old from the example above — and helps the student independently find gaps rather than pointing them out directively.

Contextual memory allows tracking which gaps the student worked on in previous sessions and suggesting a check-in after a few days — precisely the interval most effective for long-term consolidation.

An important note: the AI Mentor is a tool that strengthens the method — it doesn't perform the method for the student. The most valuable moment — the attempt to explain in your own words — must come from the student. If you start with "explain this to me" rather than "here's my understanding, check it" — the method's effect is lost.

Key Takeaways

The Feynman technique works because it switches the mode from passive reading to active recall. The feeling of understanding is an unreliable indicator. The ability to explain is a reliable one.

Four steps: write down everything you know → explain without jargon → return to the source for gaps → find an analogy and simplify.

AI strengthens each step: evaluates the understanding snapshot, asks questions like a naive listener, helps close gaps through dialogue, checks the final explanation.

The core rule: attempt first, AI second. Not the other way around.

Typical mistakes: reading before attempting, asking AI to explain at the first difficulty, skipping the analogy search, doing it only once.

Practicing the Feynman technique with AI is arguably one of the most effective combinations of proven pedagogy and a modern tool. The method existed before AI; AI makes it more complete.

---

Try the AI Mentor at opened.site — it works in Feynman mode: asks questions, checks understanding, helps find gaps. Free on the basic plan.