The Feynman Technique Is an Assessment Instrument, Not Just a Study Strategy

When a student can select the correct answer on an exam, that tells you one thing: they recognized the right option under time pressure. When a student can explain the same concept to someone who does not understand it yet, and correct that person's misunderstandings in real time, that tells you something else entirely. The Feynman Technique was designed for the second task. Assessment in higher education has largely optimized for the first.

This distinction is not a philosophical preference. It has measurable consequences for what institutions actually know about their students at the end of a course.

What the Technique Actually Demands

The Feynman Technique, named for Nobel Prize-winning physicist Richard Feynman, follows a simple structure: choose a concept, explain it as if teaching it to someone with no prior knowledge, identify where the explanation breaks down, and return to the source material to repair those gaps. The method derives its diagnostic power from the third step. Fluent recall produces confident explanations. Genuine understanding produces accurate ones. The difference is almost impossible to disguise when you are required to articulate a concept rather than recognize it.

Research on teaching-based learning consistently reinforces this. A landmark study on teachable agents, published in the Journal of Science Education and Technology, demonstrated the "protege effect": students who believed they were teaching a digital character rather than studying for themselves spent more time on learning activities and demonstrated deeper knowledge gains. The mechanism is not motivational alone. Teaching requires retrieving, organizing, and defending knowledge in ways that passive review does not. Those demands expose gaps that a confidence-based guess can conceal.

Why the Same Logic Should Apply to Assessment

Most institutions use the Feynman Technique, if at all, as a self-study recommendation. Students are told to try explaining their notes out loud before an exam. This is useful preparation, but it misses the core assessment opportunity.

The question a conceptual mastery assessment needs to answer is not "can this student explain a concept to themselves" but "can this student correct a specific misunderstanding in someone else's mental model." These are distinct tasks. The first can be performed with a rough approximation of understanding. The second requires precision.

A study published in BMC Medical Education found that over 50% of exam questions across a clinical undergraduate curriculum tested factual recall at Bloom's Level I, regardless of whether they were formatted as multiple-choice or modified essay questions. The problem is not uniquely a format problem. It is a measurement design problem: questions built to be answered rarely force the kind of articulation that reveals misconception-based evaluation opportunities.

When a student teaches a concept, the misconceptions they hold surface naturally. They mislabel a relationship, confuse a direction of causation, or correctly define a term but apply it incorrectly in a worked example. These failures are informative in ways that a wrong answer alone is not. A wrong answer tells you the student did not know the correct option. A flawed explanation tells you precisely what they believed instead.

The Field Is Beginning to Recognize This

Recent research has begun formalizing the Feynman Technique as an active assessment scaffold rather than just a private study tool. A 2025 study from arXiv introduced the Feynman Bot, an LLM-powered system designed to engage learners in question-and-answer-driven teaching conversations. Participants who used the Feynman Bot showed measurable improvements in both formative and summative assessments compared to control groups. The researchers described the approach explicitly as targeting learners "who lack peer or instructor support," using the teaching conversation itself as the assessment surface.

This is a meaningful shift. The same act of explaining that reveals understanding to the learner can, with the right instrument, reveal it to the assessor.

how Axiom Flow operationalizes this as a structured exam format

From Technique to Instrument

The gap between the Feynman Technique as self-study and the Feynman Technique as a formal teach-back assessment comes down to one design question: what is being measured, and by whom?

Oakland University's Center for Excellence in Teaching and Learning notes that the technique "facilitates deeper understanding through self-assessment and reflection" and adapts well to competency-based and case-based learning approaches. This is accurate and useful. But self-assessment and instructor-assessed performance are not the same signal. An institution that wants a reliable measure of conceptual understanding assessment cannot rely on a student's own judgment of where their explanation succeeded.

Axiom Flow goes beyond the formative assessment platform model by treating the teaching interaction itself as the scored artifact. Rather than asking whether a student can answer questions about a topic, the platform initializes an AI student with specific misconceptions drawn from the course material, then measures the quality of the student's teaching by evaluating what the AI still misunderstands afterward. The mastery score reflects not what was said, but what was understood well enough to correct.

That is the Feynman Technique functioning as an assessment higher education practitioners can act on: not a prompt to study better, but a structured instrument that captures whether understanding is present where it needs to be.

The technique has always worked this way. Most institutions simply have not built the infrastructure to use it that way. The infrastructure now exists.