From my prior experiences in teaching youth,

• I've learnt that making the subject matter relevant to their experiences is important. To do that, one would need to be a good listener, and actively ask and validate questions in order to tailor the learning experience towards something they can cognitively and emotionally identify with.
• Feedback is crucial, both for the teacher and the learner. The learner needs to be given timely, relevant and actionable feedback, and be encouraged to iterate in order to set up for the next set of feedback.
• Feedback needs to be timely, and that may not necessarily be immediately. Often, we need some time to process our thoughts, and externalize these thoughts before the salient points float to the surface. Jumping in too quickly impedes this articulation process.
• In each of the prior key points, the learner is the one setting the pace of progress. If the teacher moves too fast, the learner cannot understand the relevance and feedback given, and may believe erroneously (and prematurely) that she is correct (or wrong). Furthermore, the learner may not develop her own skillsets for metacognition in order to self-regulate in the future.
• As such, the learner needs to be given the agency to drive the progress of learning, because only the learner understands which parts do not make sense. With traditional chalk and talk classroom teaching, this is not possible. With data driven software, this may be more possible.
• The role of the teacher (or learning experience designer) in such cases, would be to, based on student data, make the experience as relatable as possible; provide feedback only when the learner requires it (even when they may not know that they do), and that includes feedback for learning processes; and allow the learner to dictate the pace of learning, in directions of their choosing.

From my experiences in AR learning experience design,