Matrixx0070 When to use
Use this to guide a learner to an answer through questions rather than telling them — a live, adaptive dialogue that surfaces their thinking, targets the actual gap, and adjusts difficulty as they respond. Best for building reasoning and durable understanding.
Not for: a one-shot explanation (use edu-explainer), a static lesson plan (use edu-lesson-plan), or when the learner explicitly just needs a fast factual answer.
Method
- Diagnose first. Ask what they're trying to do and where they're stuck; have them show their current thinking before you steer. Decision point: if the gap is a missing fact they can't derive, teach it directly, then return to questioning.
- Set the target. Pick one specific sub-goal for this exchange, not the whole topic.
- Ask, don't tell. Pose one question that moves them one step forward. Prefer questions that expose reasoning ("why did you pick that?") over yes/no.
- Adapt to the response. Correct → raise difficulty or extend. Partial → probe the shaky part. Wrong → don't correct outright; ask a question that reveals the contradiction. Decision point: after ~2 failed hints on the same step, give a worked micro-example, then resume.
- Make them articulate. Have the learner restate the idea in their own words before moving on.
- Close the loop. End with a recap in their words plus one transfer question applying it to a new case.
Example
Learner: "Why is my loop infinite?" You: "Walk me through what changes each time it runs." Learner: "Nothing — oh, i never increases." Rather than confirming, you ask: "What has to be true for a while-loop to stop?" Learner: "The condition becomes false." You: "So what would make i < 10 become false?" Learner adds i++. You raise difficulty: "What if the step were i += 3 — would it still terminate?" Recap in their words + transfer to a countdown loop.
Pitfalls
- Giving the answer at the first sign of struggle, so the learner never builds reasoning.
- Asking vague "does that make sense?" checks instead of questions that reveal thinking.
- Stacking multiple questions at once, overwhelming instead of guiding one step.
- Never escalating out of Socratic mode even after repeated failure — hint fatigue; drop to a worked example.
Output format
Session goal: <what the learner wants> | Level: <...>
Turn-by-turn:
- Diagnose: <opening question + learner's shown thinking>
- Q1 → (learner response) → adapt: <correct/partial/wrong branch>
- Q2 → ...
- [Fallback used?]: worked micro-example when stuck
Learner's own-words recap: <...>
Transfer question (new case): <...>
Next step / what to practice: <...>