Teachers often ask students things like this: “Did you understand?”, “Do you have any questions?”, or “Who can answer this question?”, and so on. The problem with this kind of questions is that the teacher is giving the student control over the direction of the lecture, and, quite often, the students are very bad at making the lecture interesting or useful (unless the audience is a small number of extremely interested and cooperative researchers).

Depending on the students, the goal of every lecture is to successfully communicate some ideas. However, sometimes, the goal from students’ perspective is to just finish the lecture as soon as possible, pass the test, get the degree, get a job, and move on. This is typically due to having different priorities, which is not bad per se. For example, if the student have been stuck for a long time fighting a tough boss (tougher than the lecture) in an interesting video game (more interesting than the lecture), they might be simulating and brainstorming strategies to beat this boss during the lecture, which is in addition to—unsuccessfully, of course—trying to understand the material of the lecture. This is the first point to keep in mind: students may be present, but their minds may be absent. They should be in the context of the lecture and paying good attention to be able to comprehend the material.

Now, coming from the student’s perspective, when they say “Yes, I understand”, they probably don’t understand in the lecturer’s sense. They understand in their (probably naive) sense. They might have understood the subject to the extent that enables them to barely pass the test, which is a really bad measure of understanding. A key measure of understanding is the ability to explain (i.e., the Feynman definition of understanding). So, instead of asking “Do you understand?”, you can ask “Can you explain?”

A teacher should not let the students decide whether things are going well. Whenever a teacher asks a question, they should ask it not in the sense that they are proof-checking their answers and trying to beat them into failure, but in the sense that they want them to voluntarily contribute their answers to the lecture. For example, a good teacher can guide the students along an seemingly-straightforward intuition until a contradiction or a missing piece prompts the students to voluntarily ask questions. This will make them remember the concept from the memory of this internal conflict. Usually, the edge cases are what makes a particular object or phenomenon interesting and unique. Questions are a good way to test student’s understanding as well, especially open-ended ones. Let them ask interesting questions and let them find the answers by themselves, guiding them only when necessary by intuiting from their perspectives.

One might ask: “Why teach students this way? Why not let them brainstorm strategies to beat the final boss?”, which is valid, but not in an educational setting. The main problem is that the teacher is paid to do the job of teaching given a time and location constraint (e.g., lecture), among others. If the teacher is willing to do their job effectively, which is to teach; or to convey knowledge to students and equip them with the tools to practice it in the real world, then I argue that allowing the students to comprehend such that they can explain the unique characteristics of such knowledge is a robust technique for ensuring that they have developed a good model of it. If you’re deep learning-savvy, you can sort of think of it as training and testing an autoencoder.