Stop explaining, start scaffolding

Effective learning often suffers from an excess of clarity. When researchers presented children with a toy, those who observed an accidental demonstration explored more functions than those who received a deliberate explanation (The Playbook). Direct instruction, while efficient for rote tasks like CPR, often fails to foster the deep, conceptual understanding required for complex problem-solving (The Argument). This gap between memorizing procedures and grasping underlying principles is central to the concept of transfer, or the ability to apply knowledge in novel situations (The Science of Creativity). Instead of providing immediate answers, educators should design experiences that allow learners to encounter questions that reveal their own gaps in understanding (Syd Malaxos A public high school STEM…). By monitoring the length of your own explanations, you create the necessary space for discovery, ensuring that the learner remains responsible for the reasoning process rather than merely repeating provided information (Syd Malaxos A public high school STEM…).

“The children played with the toy longer, tried more different actions, and discovered more of the ‘hidden’ features when the experimenter squeaked the beeper accidentally than they did when she deliberately tried to teach them.”

The Playbook

AI tools frequently undermine this process by being too helpful. The new TutorMoments benchmark, which tested 462 math tutoring transcripts, found that frontier models often provide hints when students would benefit more from productive struggle (Edtech Insiders). A randomized trial involving nearly 1,000 students showed that those using standard AI chat interfaces performed 17 percent worse on exams after access to the tool was removed, whereas students who used versions with teacher-designed safeguards showed no significant difference from those without AI (Syd Malaxos A public high school STEM…). To avoid this dependency, you must ensure that learners bring their own attempts and questions to the AI, using the tool to extend their existing reasoning rather than replacing it. This requires shifting from an AI optimized for correct answers to one that prompts students to defend their interpretations or explain the mechanics behind a result.

True learning is a co-creative process that requires moving beyond simple activity labels. The learning pyramid, which assigns fixed retention percentages to methods like lecturing or practicing, lacks empirical foundation and ignores the quality of the learner's thinking (The Collaboration Chronicle). Whether using a lecture or an AI tool, the effectiveness depends on whether the student is actively elaborating on the material through strategies like self-explanation. Since every learner enters through a different door—some preferring definitions, others needing to see the mechanism or the real-world application—you should identify their starting point before pushing them through the other perspectives. By focusing on ownership of the idea, you move the student from merely remembering a definition to applying the concept in a situation they have never encountered before (The Divergent Talent Alchemist).