The Feynman Technique with AI

Explaining a concept to someone else can expose a gap that rereading leaves hidden. Research on self-explanations, including work by Chi and colleagues on learning from worked examples, studies how learners generate explanations while solving problems. The popular Feynman Technique label is a study heuristic rather than one standardized experimental protocol, so avoid claiming a universal percentage improvement or attributing a precise four-step recipe to a particular study. The practical cycle is to choose a topic, explain it from memory, locate uncertain parts, check a trusted source and try again. An AI partner can ask follow-up questions such as 'why does that step follow?' or 'what happens in this edge case?' Set a rule that it should question the learner before giving a complete explanation. If the model writes the polished paragraph first, the learner may recognize the words without being able to reproduce the reasoning. Short, plain language is useful when it preserves the essential mechanism; oversimplifying a necessary condition can make the explanation false. Keep technical terms when they carry meaning, then define them with examples. After a first explanation, compare it with the course text or primary reference. Mark each factual claim as supported, incomplete or wrong. Ask AI for a counterexample or an alternative representation, but verify its challenge as well. Revise the explanation in the learner’s own words. A diagram, equation or small worked example can reveal where a verbal account hides a missing step. Finally, test transfer. Can the student use the idea in a new problem, answer a skeptical question or recognize when it does not apply? A fluent speech alone may mask shallow understanding. Respect the course rules about AI and do not upload confidential material. The tool’s role is to be a persistent, critical listener; the student’s role is to construct and defend the explanation.