How can you tell a correct answer from understanding?

What does explaining an answer to a peer reveal?

Collaborative learning gains a concrete mechanism when students have to explain their reasoning to one another. Agreement on an answer alone tells the teacher less.

Why do experts miss the step a beginner needs?

Familiarity can hide prerequisites. Compare the techniques each speaker uses to expose the gap between what a presenter assumes and what a learner can reconstruct.

Hide a corner instead of asking for three rules

Put 2, 3 and 8 at the corners of a triangle: 2 cubed equals 8. Hide the 8 to ask for a power, the 3 to ask for a logarithm, or the 2 to ask for a root. Ask the learner to name the missing quantity before calculating. Someone who knows the calculator buttons but not which question they answer now has a way to express the confusion.

A tougher transfer test is to rewrite a probability puzzle without changing its sampling rule. In the sphere puzzle, choose lines through the center, then flip fair coins to choose their endpoints. Now one can count choices instead of computing the areas of changing spherical triangles. Why do these steps still produce uniformly random points? “Use coins when you see a sphere” misses the part of the explanation that matters.

Does getting the answer mean understanding the concept?

Concept questions in physics and geometric intuition in linear algebra reveal gaps that procedural success can hide. Neither approach rules out calculation practice.

After the definition, ask for a next step

Checking whether y² is near x and writing a procedure to find a square root are different tasks in MIT’s example. Give a learner a small robot that can compare, update a guess and repeat. What instruction would eventually stop it? Stanford teaches programming in a constrained robot world and considers the clarity of the program as well as the result. Hofstadter interprets categorization through analogy. To test transfer, change the task and ask what plays the role of the old guess and its stopping condition.

After a mistake turns red, what changes for the learner?

Put one learner in three of Khan’s classroom scenes. Mastery learning gives the learner more time. A dashboard helps a teacher find a peer tutor. In the AI demonstration the tutor asks for the next step before identifying a possible error in distribution. All three could follow the same red mark; what matters is the next action. Give the learner a different-looking problem that still requires distributing a negative sign and ask for an independent explanation. The three scenes are Khan’s accounts, not separate evaluations of outcomes.

How do you carry a computed answer back to the world?

Wolfram places computation between formulation and verification. In Stoll’s sound-speed demonstration the learner must identify what frequency the counter reads and how far to move the microphone to see the phase repeat. Kay’s students go the other way: program repeated increments of speed and distance, then measure displacements in equally spaced frames of a fall. Ask which observation would make them withdraw the claim of constant acceleration or reject the supposed wavelength.

Hand the explanation to someone else to reconstruct

Knuth asks a reader to understand an algorithm in both formal and everyday language. Howard's scheduler notebook makes that comparison executable: change a parameter, predict the output from the prose, then run the implementation and its tests. Matuschak offers a parallel distinction between recognizing a word on a card and retrieving it in a sentence. Cover the result and give another reader a new input or context. Ask for a prediction and its reasoning before uncovering the worked example.

Filling in a pronoun and explaining its referent

Lenat asks what a writer expects a reader to supply about a pronoun or an omitted event between sentences. That suggests a small test of Sutskever's hypothesis. Keep the wording nearly fixed but swap a role so the natural referent changes. Record the model's judgment in both cases and ask which fact changed the reference. A fluent continuation and a pair of judgments that move with the facts supply different evidence.

One candle, several observation tasks

Following the wax directs attention to the wick. Following warm air directs it to convection. The Faraday reconstructions offer a commentary track that helps the viewer change questions while keeping the object fixed. Peterson’s comparison of theories to tools suggests a comprehension check: state the assumption doing the work, then predict what a different observation task would require.

An answer given aloud and one given by hand

In Grey’s account of a split-brain task, the participant can report seeing nothing while a manual response selects the object. An audio-only record could turn a failure to name into a failure to recognize. That observation is enough to change the test sheet: record the response channel along with the answer.

What survives when the notes are closed?

In an introductory lecture Peterson contrasts highlighting with attempted recall and discusses spacing. Turn the advice into a small test: close the notes, write a definition, then reconstruct it with a new example tomorrow. A separate personality lecture discusses predictive validity, prompting a further question: what does each score predict? Recognizing a familiar sentence and transferring a concept can receive identical marks for different performances.

What can a pause button do for a lesson?

Grey moves from a classroom’s pace to media a learner can stop. Pausing changes timing; understanding can be probed the next day with a case in a different setting. Peterson’s suggestion to close the materials and attempt recall offers a further check. Compare where learners use pause and what they can subsequently do on their own.