Back to Skills

socratic-teaching-scaffolds

lyndonkl
Updated Today
26 views
5
5
View on GitHub
Designaidesign

About

This skill uses Socratic questioning to guide learners through complex concepts rather than providing direct answers. It helps users discover insights by asking probing questions, revealing contradictions in thinking, and providing scaffolded learning frameworks. Developers should activate it for teaching technical concepts, onboarding team members, or when users request explanations using phrases like "help me understand" or "teach me."

Documentation

Socratic Teaching Scaffolds

Table of Contents

  1. Purpose
  2. When to Use
  3. What Is It
  4. Workflow
  5. Socratic Question Types
  6. Scaffolding Levels
  7. Common Patterns
  8. Guardrails
  9. Quick Reference

Purpose

Socratic Teaching Scaffolds guide learners to discover knowledge through strategic questioning and progressive support removal. This skill transforms passive explanation into active discovery, corrects misconceptions by revealing contradictions, and builds durable understanding through self-generated insights.

When to Use

Invoke this skill when you need to:

  • Teach complex concepts where understanding beats memorization (algorithms, scientific theories, philosophical ideas)
  • Correct deep misconceptions that resist direct explanation (statistical fallacies, physics intuitions, programming mental models)
  • Help learners develop problem-solving skills, not just solutions
  • Design learning experiences that build from concrete to abstract understanding
  • Onboard professionals through guided discovery rather than documentation dumps
  • Create self-paced learning materials with built-in feedback loops
  • Mentor through questions that develop independent thinking
  • Bridge expertise gaps across different knowledge levels

User phrases that trigger this skill:

  • "Teach me [concept]"
  • "Help me understand [topic]"
  • "Explain like I'm [expertise level]"
  • "I don't get why [misconception]"
  • "What's the best way to learn [skill]?"
  • "How should I think about [concept]?"

What Is It

A teaching framework combining Socratic questioning (strategic questions that guide discovery) with instructional scaffolding (temporary support that fades as competence grows).

Core components:

  1. Question Ladders: Sequences from simple to complex that build understanding incrementally
  2. Misconception Detectors: Questions that reveal faulty mental models through contradiction
  3. Feynman Explanations: Build-up from simple analogies to technical precision
  4. Worked Examples with Fading: Full solutions → partial solutions → independent practice
  5. Cognitive Apprenticeship: Model thinking process explicitly, then transfer to learner

Quick example (Teaching Recursion):

Question Ladder:

  1. "Can you break this problem into a smaller version of itself?" (problem decomposition)
  2. "What would happen if we had only one item?" (base case discovery)
  3. "If we could solve the small version, how would we use it for the big version?" (recursive case)
  4. "What prevents this from running forever?" (termination reasoning)

Misconception Detector:

  • "Will this recursion ever stop? Trace it with 3 items." (reveals infinite recursion misunderstanding)

Feynman Progression:

  • Level 1: "Like Russian nesting dolls—each contains a smaller version"
  • Level 2: "Function calls itself with simpler input until base case"
  • Level 3: "Recursive definition: f(n) = g(f(n-1), n) with f(0) = base"

Workflow

Copy this checklist and track your progress:

Socratic Teaching Progress:
- [ ] Step 1: Diagnose learner's current understanding
- [ ] Step 2: Design question ladder and scaffolding plan
- [ ] Step 3: Guide discovery through questioning
- [ ] Step 4: Fade scaffolding as competence grows
- [ ] Step 5: Validate understanding and transfer

Step 1: Diagnose learner's current understanding

Ask probing questions to identify current knowledge level, misconceptions, and learning goals. See Socratic Question Types for diagnostic question categories.

Step 2: Design question ladder and scaffolding plan

Build progression from learner's current state to target understanding. For straightforward teaching → Use resources/template.md. For complex topics with multiple misconceptions → Study resources/methodology.md.

Step 3: Guide discovery through questioning

Ask questions in sequence, provide scaffolding (hints, worked examples, analogies) as needed. See Scaffolding Levels for support gradations. Adjust based on learner responses.

Step 4: Fade scaffolding as competence grows

Progressively remove hints, provide less complete examples, ask more open-ended questions. Monitor for struggle (optimal challenge) vs frustration (too hard). See resources/methodology.md for fading strategies.

Step 5: Validate understanding and transfer

Test with novel problems, ask for explanations in learner's words, check for misconception elimination. Self-check using resources/evaluators/rubric_socratic_teaching_scaffolds.json. Minimum standard: Average score ≥ 3.5.

Socratic Question Types

1. Clarifying Questions (Understand current thinking)

  • "What do you mean by [term]?"
  • "Can you give me an example?"
  • "How does this relate to [known concept]?"

2. Probing Assumptions (Surface hidden beliefs)

  • "What are we assuming here?"
  • "Why would that be true?"
  • "Is that always the case?"

3. Probing Reasons/Evidence (Justify claims)

  • "Why do you think that?"
  • "What evidence supports that?"
  • "How would we test that?"

4. Exploring Implications (Think through consequences)

  • "What would happen if [change]?"
  • "What follows from that?"
  • "What are the edge cases?"

5. Questioning the Question (Meta-cognition)

  • "Why is this question important?"
  • "What are we really trying to understand?"
  • "How would we know if we understood?"

6. Revealing Contradictions (Bust misconceptions)

  • "Earlier you said [X], but now [Y]. How do these fit?"
  • "If that's true, why does [counterexample] happen?"
  • "What would this predict for [test case]?"

Scaffolding Levels

Provide support that matches current need, then fade:

Level 5: Full Modeling (I do, you watch)

  • Complete worked example with thinking aloud
  • Explicit strategy articulation
  • All steps shown with rationale

Level 4: Guided Practice (I do, you help)

  • Partial worked example
  • Ask learner to complete steps
  • Provide hints before errors

Level 3: Coached Practice (You do, I help)

  • Learner attempts independently
  • Intervene with questions when stuck
  • Guide without giving answers

Level 2: Independent with Feedback (You do, I watch)

  • Learner solves alone
  • Review and discuss afterwards
  • Identify gaps for next iteration

Level 1: Transfer (You teach someone else)

  • Learner explains to others
  • Learner creates examples
  • Learner identifies misconceptions in others

Fading strategy: Start at level matching current competence (not Level 5 by default). Move down one level when learner demonstrates success. Move up one level if learner struggles repeatedly.

Common Patterns

Pattern 1: Concept Introduction (Concrete → Abstract)

  • Start: Real-world analogy or example
  • Middle: Formalize with terminology
  • End: Abstract definition with edge cases
  • Example: Teaching pointers (address on envelope → memory location → pointer arithmetic)

Pattern 2: Misconception Correction (Prediction → Surprise → Explanation)

  • Ask learner to predict outcome
  • Show actual result (contradicts misconception)
  • Guide discovery of correct mental model
  • Example: "Will this float? [test with 0.1 + 0.2 in programming] Why not exactly 0.3?"

Pattern 3: Problem-Solving Strategy (Model → Practice → Reflect)

  • Model strategy on simple problem (think aloud)
  • Learner applies to similar problem (with coaching)
  • Reflect on when strategy applies/fails
  • Example: Teaching debugging (print statements → breakpoints → hypothesis testing)

Pattern 4: Depth Ladder (ELI5 → Undergraduate → Expert)

  • Build multiple explanations at different depths
  • Let learner choose starting point
  • Provide "go deeper" option at each level
  • Example: Teaching neural networks (pattern matching → weighted sums → backpropagation → optimization theory)

Pattern 5: Discovery Learning (Puzzle → Hints → Insight)

  • Present puzzling phenomenon or problem
  • Provide graduated hints if stuck
  • Guide to "aha" moment of discovery
  • Example: Teaching recursion (Towers of Hanoi → break into subproblems → recursive solution)

Guardrails

Zone of proximal development:

  • Too easy = boredom, too hard = frustration
  • Optimal: Can't do alone, but can with guidance
  • Adjust scaffolding level based on struggle signals

Don't fish for specific answers:

  • Socratic questioning isn't a guessing game
  • If learner's reasoning is sound but reaches different conclusion, explore their path
  • Multiple valid approaches often exist

Avoid pseudo-teaching:

  • Don't just ask questions without purpose
  • Each question should advance understanding or reveal misconception
  • If question doesn't help, provide direct explanation

Misconception resistance:

  • Deep misconceptions resist single corrections
  • Need multiple exposures to contradictions
  • May require building correct model from scratch before dismantling wrong one

Expertise blind spots:

  • Experts forget what was hard as beginners
  • Make implicit knowledge explicit
  • Slow down automated processes to show thinking

Individual differences:

  • Some learners prefer exploration, others prefer structure
  • Adjust scaffolding style to learner preferences
  • Monitor for frustration vs productive struggle

Quick Reference

Resources:

5-Step Process: Diagnose → Design Ladder → Guide Discovery → Fade Scaffolding → Validate Transfer

Question Types: Clarifying, Probing Assumptions, Probing Evidence, Exploring Implications, Meta-cognition, Revealing Contradictions

Scaffolding Levels: Full Modeling → Guided Practice → Coached Practice → Independent Feedback → Transfer (fade progressively)

Patterns: Concrete→Abstract, Prediction→Surprise→Explanation, Model→Practice→Reflect, ELI5→Expert, Puzzle→Hints→Insight

Guardrails: Zone of proximal development, purposeful questions, avoid pseudo-teaching, resist misconceptions, make implicit explicit

Quick Install

/plugin add https://github.com/lyndonkl/claude/tree/main/socratic-teaching-scaffolds

Copy and paste this command in Claude Code to install this skill

GitHub 仓库

lyndonkl/claude
Path: skills/socratic-teaching-scaffolds

Related Skills

sglang

Meta

SGLang is a high-performance LLM serving framework that specializes in fast, structured generation for JSON, regex, and agentic workflows using its RadixAttention prefix caching. It delivers significantly faster inference, especially for tasks with repeated prefixes, making it ideal for complex, structured outputs and multi-turn conversations. Choose SGLang over alternatives like vLLM when you need constrained decoding or are building applications with extensive prefix sharing.

View skill

evaluating-llms-harness

Testing

This Claude Skill runs the lm-evaluation-harness to benchmark LLMs across 60+ standardized academic tasks like MMLU and GSM8K. It's designed for developers to compare model quality, track training progress, or report academic results. The tool supports various backends including HuggingFace and vLLM models.

View skill

llamaguard

Other

LlamaGuard is Meta's 7-8B parameter model for moderating LLM inputs and outputs across six safety categories like violence and hate speech. It offers 94-95% accuracy and can be deployed using vLLM, Hugging Face, or Amazon SageMaker. Use this skill to easily integrate content filtering and safety guardrails into your AI applications.

View skill

langchain

Meta

LangChain is a framework for building LLM applications using agents, chains, and RAG pipelines. It supports multiple LLM providers, offers 500+ integrations, and includes features like tool calling and memory management. Use it for rapid prototyping and deploying production systems like chatbots, autonomous agents, and question-answering services.

View skill