How to teach prompt engineering to 6th graders

 6th Grade Prompt Engineering Course

A Complete Guide to Talking with AI

Food for Thought: Why This Matters More Than Ever

Imagine if your grandparents had to learn how to use the internet when they were your age. That's exactly what's happening to you right now with AI! Just like learning to type replaced learning perfect handwriting, and using search engines became more important than memorizing encyclopedia facts, prompt engineering is becoming the new essential writing skill.

In the age of agentic AI - where AI assistants can use tools, make decisions, and complete complex tasks independently - students who can communicate clearly with AI will have superpowers that previous generations couldn't even imagine. This isn't just about homework help anymore. It's about learning to collaborate with AI partners that can:

• Research and analyze information faster than any human
• Create presentations, art, and code in seconds
• Solve complex problems by breaking them into manageable steps
• Access real-time information from across the internet
• Help you learn any subject at your own pace

The students who master prompt engineering today will be the leaders, innovators, and problem-solvers of tomorrow. They'll know how to turn AI from a simple question-answering tool into a powerful thinking partner. While others struggle to get basic responses, prompt engineering experts will orchestrate AI systems to tackle challenges we haven't even discovered yet.

This course isn't just preparing you for middle school or high school - it's preparing you for a world where human-AI collaboration is as natural as using a smartphone is today. The sooner you start, the bigger advantage you'll have in this AI-powered future.

How to teach prompt engineering to 6th graders

AI communication curriculum for middle school

Prompt engineering lesson plans elementary grades

Children learning AI prompt conversation skills

6th-grade artificial intelligence education

A Complete Guide to Talking with AI

Course Overview

Welcome to the exciting world of Prompt Engineering! In this course, you'll learn how to communicate effectively with AI systems like ChatGPT, Claude, and other smart computer programs. Think of it like learning a special language that helps you get exactly what you want from AI helpers.

---

Unit 1: Introduction to Prompts and AI Communication

What is a Prompt?

A prompt is like giving instructions to a very smart robot. Just like when you ask a friend to help you with homework, you need to be clear about what you want. The better your instructions, the better help you'll get!

Example of a simple prompt: "Help me write a story about a dragon who loves pizza."

https://www.youtube.com/shorts/aAvrSbDOFtU

Types of Prompts

User Prompt vs System Prompt

• User Prompt: This is what YOU type to the AI. It's like talking directly to your AI friend.
• System Prompt: This is like giving the AI a special job description before you start talking. It tells the AI how to behave (like "You are a helpful math tutor").

What are Tokens?

Think of tokens like LEGO blocks that make up words. When you type "Hello world", the AI sees it as small pieces:

• "Hello" might be 1 token
• "world" might be 1 token
• So "Hello world" = 2 tokens

The AI has a limit on how many LEGO blocks (tokens) it can work with at once, just like you might have a limit on how many LEGO pieces you can use in one building session.

---

Unit 2: Prompt Structure and Flow

Task-First vs Task-Last Structure

Task-First Structure (Recommended)

TASK: Write a poem about friendship
DETAILS: Make it rhyme and include animals
EXAMPLE: Like "Cats and dogs can be best friends..."

Task-Last Structure

I want you to think about friendship and animals.
Consider how they might get along together.
Now write a poem about friendship that rhymes and includes animals.

The Anatomy of a Great Prompt

A great prompt has these parts (think of it like a sandwich):

1. Context (The bread) - Background information
2. Task (The main filling) - What you want done
3. Format (The toppings) - How you want the answer
4. Examples (The sauce) - Shows what "good" looks like

---

Unit 3: In-Context Learning and Shot Types

What is In-Context Learning?

This is like showing the AI examples before asking it to do something new. It's like showing your little brother how to tie shoes before asking him to tie his own.

Types of "Shots"

Zero-Shot

No examples given. You just ask the AI to do something cold.

"Write a haiku about summer."

One-Shot

You give ONE example first.

"Here's a haiku about winter:
Snow falls gently down
Covering the sleeping earth
Winter's quiet gift

Now write a haiku about summer."

Few-Shot

You give a FEW examples (usually 2-5).

"Here are some haikus:

Winter haiku:
Snow falls gently down
Covering the sleeping earth
Winter's quiet gift

Spring haiku:
Cherry blossoms bloom
Painting the world soft pink
New life awakens

Now write a haiku about summer."

Many-Shot (20+ examples)

You give MANY examples. This is like showing someone 20 different ways to solve math problems before asking them to solve a new one.

---

Unit 4: Advanced Reasoning Techniques

Chain of Thought (CoT)

This is like asking the AI to "show its work" in math class.

Regular prompt: "What's 15% of 80?"

Chain of Thought prompt: "What's 15% of 80? Please show your step-by-step thinking."

Tree of Thought

This is like having the AI consider multiple paths to solve a problem, like a choose-your-own-adventure book.

Example: "I need to plan a class party. Think through three different approaches:

1. Indoor activities
2. Outdoor activities
3. Mixed indoor/outdoor For each approach, consider the pros and cons, then recommend the best option."

Giving AI Time to Think

Add phrases like:

• "Take your time to think through this"
• "Let's work through this step by step"
• "First, let me think about..."

---

Unit 5: Using Judge Models and Error Checking

What is a Judge Model?

A judge model is like having a teacher check your work. You can ask the AI to double-check its own answers.

Example:

"Solve this math problem: 25 × 4 = ?
Now, check your answer by using a different method."

Self-Evaluation Prompts

"Write a short story about space exploration.
After you write it, please:
1. Check if it has a clear beginning, middle, and end
2. Make sure the science facts are correct
3. Suggest one improvement"

---

Unit 6: Best Practices for Effective Prompt Writing

The CLEAR Framework

• Concrete: Be specific about what you want
• Length: Specify how long the response should be
• Examples: Show what good looks like
• Audience: Say who this is for
• Role: Tell the AI what expert to be

Persona Prompts

Make the AI pretend to be an expert!

Example: "You are a marine biologist who loves teaching kids about ocean life. Explain how dolphins communicate in a way that's fun and easy to understand for 6th graders."

Getting Better Examples

To get the AI to give you better examples:

1. Ask for multiple options: "Give me 5 different ways to..."
2. Specify the type: "Give me creative examples" or "Give me realistic examples"
3. Ask for improvements: "Make these examples more interesting"

---

Unit 7: Response Formats and Outputs

Specifying Output Format

Always tell the AI exactly how you want the answer formatted.

Text Formats

• Paragraph: "Write this as one paragraph"
• List: "Give me a numbered list"
• Table: "Put this information in a table with columns for..."
• Outline: "Create an outline with main points and sub-points"

Creative Formats

• Story: "Write this as a short story"
• Poem: "Make this into a poem that rhymes"
• Dialogue: "Write this as a conversation between two people"
• Letter: "Format this as a friendly letter"

Structured Formats

• Step-by-step: "Give me numbered steps"
• Pros and cons: "List the advantages and disadvantages"
• Compare and contrast: "Show similarities and differences"

Length Specifications

• "In 50 words"
• "Write 3 sentences"
• "Give me a one-page explanation"
• "Keep it under 100 words"

---

Unit 8: Breaking Down Complex Tasks

The Chunking Strategy

Big tasks are like eating a whole pizza - you need to cut it into slices!

Instead of: "Plan a school science fair"

Try this: "Let's plan a school science fair step by step: Step 1: What are the main categories of science projects we should include? Step 2: How should we organize the space and timing? Step 3: What supplies and materials will we need? Step 4: How can we make it fun for visitors?

Let's start with Step 1."

Sequential Prompting

After getting the first answer, build on it:

1. First prompt: Get the basic structure
2. Second prompt: Add details to each part
3. Third prompt: Refine and improve

---

Unit 9: Advanced Reasoning and Self-Critique

Iterative Improvement

Teach the AI to improve its own work:

"Write a persuasive paragraph about why students should recycle.
Now, read what you wrote and improve it by:
1. Adding one more strong reason
2. Making the language more exciting
3. Including a call to action"

Tree of Thought Analysis

"I need to choose between three science fair projects:
1. Volcano eruption
2. Solar system model
3. Plant growth experiment

For each option, analyze:
- How interesting it would be
- How difficult it would be
- What materials I'd need
- How long it would take

Then recommend which one I should choose and why."

---

Unit 10: Agentic AI and Autonomous Tool Use

What is Agentic AI?

Agentic AI is like having a super-smart assistant that can use different tools to help you, just like you might use a calculator for math, a dictionary for words, and a map for directions.

What are Tools?

Tools are special programs the AI can use:

• Search tool: Like Google, but for the AI
• Calculator: For complex math
• Image creator: To make pictures
• Code runner: To test computer programs
• File reader: To read documents you upload

How Agentic AI Works

1. You give it a big goal
2. It breaks the goal into smaller steps
3. It decides which tools to use for each step
4. It uses the tools automatically (autonomously)
5. It puts all the results together

Prompting for Tool Use

"I need to research endangered animals and create a presentation.
Please:
1. Search for information about 5 endangered animals
2. Find current population numbers
3. Create a simple chart showing the data
4. Suggest images that would work well in a presentation"

---

Unit 11: Model Comparison and Selection

Different AI Models

Different AI models are like different types of vehicles:

• GPT-4: Like a luxury car - great for complex tasks
• Claude: Like a reliable truck - good for analysis and writing
• Smaller models: Like bicycles - faster but less powerful

Choosing the Right Model

• For creative writing: Models good with language
• For math problems: Models good with calculations
• For research: Models with search capabilities
• For coding: Models trained on programming

---

Complete Glossary for 6th Graders

A

Agentic AI: AI that can work independently and use different tools to complete tasks, like a robot assistant that can use a calculator, search the internet, and write reports all by itself.

API (Application Programming Interface): A way for different computer programs to talk to each other. Think of it like a universal translator that helps different apps share information.

C

Chain of Thought: A way of prompting AI to show its step-by-step thinking, like showing your work in math class.

Chatbot: A computer program that can have conversations with people through text, like texting with a very smart robot.

Context: Background information you give to AI to help it understand what you're asking for.

F

Few-Shot Learning: Teaching AI by showing it a few examples before asking it to do something new.

I

In-Context Learning: Teaching AI by giving it examples and information within your conversation, like showing someone how to do something right before asking them to try it.

J

Judge Model: An AI system that checks and evaluates other AI responses for accuracy and quality, like a teacher grading homework.

L

Large Language Model (LLM): A very smart computer program trained on lots of text that can understand and generate human-like language. Think of it as a computer that has read millions of books and can write like a human.

M

MCP (Model Context Protocol): A system that helps AI models connect with and use different tools and databases, like a universal remote control for AI.

O

One-Shot Learning: Teaching AI by showing it just one example before asking it to do something similar.

P

Persona: A character or role you ask the AI to pretend to be, like asking it to act like a scientist or a storyteller.

Prompt: Instructions or questions you give to an AI system to get it to do what you want.

Prompt Engineering: The skill of writing really good prompts to get the best results from AI systems.

S

System Prompt: Special instructions that tell the AI how to behave throughout an entire conversation, like giving it a job description.

T

Token: The smallest unit of text that an AI processes, like individual puzzle pieces that make up words and sentences.

Tree of Thought: A method where AI considers multiple different approaches to solving a problem, like exploring different paths in a maze.

U

User Prompt: The specific question or instruction that you type to the AI during your conversation.

Z

Zero-Shot Learning: Asking AI to do something without giving it any examples first, like asking someone to solve a puzzle they've never seen before.

---

Final Project Ideas

1. Create a Perfect Prompt: Design a prompt that gets an AI to help you plan your ideal birthday party.

2. Build a Prompt Chain: Create a series of connected prompts that help you research and write a report on your favorite animal.

3. Design an AI Assistant: Write prompts that would make an AI act like the perfect homework helper for your grade level.

4. Prompt Comparison: Take the same task and write it three different ways (zero-shot, few-shot, and with chain of thought) and see which works best.

5. Create a Glossary: Build your own glossary of AI terms using prompts to help you understand and explain each concept.

---

Assessment Rubric

Students will be evaluated on:

• Understanding of concepts (25%)
• Quality of prompts written (25%)
• Ability to break down complex tasks (25%)
• Creative application of techniques (25%)

Remember: The goal isn't to trick the AI, but to communicate clearly and effectively to get the best help possible!

RAG Indexing: Like Building a Super Smart Library

What is RAG?

Imagine you have a really smart robot friend who can answer almost any question. But sometimes, this robot doesn't know about your specific school, your family, or the latest news. RAG (Retrieval-Augmented Generation) is like giving your robot friend access to a super organized library so it can look up information and give you better answers.

What is RAG Indexing?

RAG indexing is like organizing a huge library so that information can be found super quickly. Instead of having books scattered everywhere, we organize them in a special way that makes finding the right information lightning fast.

How Does It Work?

Step 1: Breaking Things Down (Chunking)

Think of a really long book. Instead of trying to remember the whole book at once, we break it into smaller chapters or even paragraphs. This is called "chunking." It's like cutting a giant pizza into slices - much easier to handle!

Step 2: Creating a Secret Code (Embeddings)

Each chunk of information gets turned into a special code made of numbers. Think of it like giving each piece of information a unique fingerprint. These number codes help a computer understand what each piece of text is really about, not just the words themselves.

For example, the chunks "dogs are pets" and "cats are animals" would get similar number codes because they're both about pets, even though they use different words.

Step 3: Building the Super Library (Vector Database)

All these number codes get stored in a special kind of database called a "vector database." Think of it like a magical library where instead of organizing books by alphabet, we organize them by how similar their "fingerprints" are. Books about dogs would be near books about cats, and books about math would be near books about science.

How Do We Find Information?

Step 1: Turn Your Question into a Code

When you ask a question, the computer turns your question into the same type of number code.

Step 2: Find the Best Matches

The computer looks through all the stored codes to find the ones that are most similar to your question's code. It's like finding all the library books that have fingerprints similar to what you're looking for.

Step 3: Give the Smart Robot the Right Information

The computer takes the best matching information and gives it to the smart robot, so the robot can use that specific information to answer your question really well.

Why is This Awesome?

Better Answers: Instead of guessing, the robot can look up the exact information you need and give you accurate answers.

No Made-Up Stuff: Sometimes smart robots make up answers when they don't know something (like making up a fake book title). With RAG, the robot uses real information from the library, so it's less likely to make things up.

Always Up-to-Date: When new information comes in, we can add it to our special library without having to teach the robot everything all over again.

Super Fast: The special way we organize information makes finding the right stuff incredibly quick - like having a librarian who can instantly teleport to the exact book you need.

Real-Life Example

Let's say you ask, "What's my school's mascot?" Without RAG, the robot might not know. But with RAG indexing, if information about your school is in the organized library, the robot can quickly find that information and tell you, "Your school's mascot is the Eagles!"

RAG indexing is basically like giving any smart computer system a perfectly organized, searchable library so it can always find the best information to help answer your questions.

Building a Montessori Math AI: 6-Day Capstone Project

Creating Visual Anchor Charts for Math Word Problems

Project Overview

Students will create an AI system that takes math word problems and generates visual anchor charts showing step-by-step solutions using Montessori manipulatives. The system will use the "Read, Build, Draw, Write" process to create educational visual guides.

Montessori Color Code Reference

• Green Units (1s): Individual units
• Blue Units (10s): Ten bars
• Red Units (100s): Hundred squares
• Green Units (1000s): Thousand cubes

---

Day 1: Understanding the Problem & RAG Foundations

Learning Objectives

Students will understand what makes a good visual math explanation and learn how AI needs organized information to work properly.

Activities

Morning Session: Analyzing Great Math Explanations

1. Anchor Chart Analysis (30 minutes)
• Show students 5 different anchor charts for the same math problem
• Have them rank from "most helpful" to "least helpful"
• Discuss what makes a visual explanation clear and useful
2. Montessori Method Review (45 minutes)
• Practice solving 3 word problems using physical Montessori materials
• Document each step with photos
• Write down the exact sequence: Read → Build → Draw → Write

Afternoon Session: Introduction to AI Training

3. AI Needs Data (30 minutes)
• Explain: "AI is like a student who learns from examples"
• Show how humans need many examples to learn patterns
• Introduce the concept: "To teach AI, we need to give it lots of good examples"
4. Building Our Training Database (45 minutes)
• Create a shared digital folder structure:
• Word Problems (organized by operation type)
• Step-by-Step Solutions
• Visual Examples
• Common Mistakes to Avoid

Homework

Each student brings in 2 word problems from their math textbook (different operation types)

---

Day 2: Creating Training Data & Understanding Context

Learning Objectives

Students will create high-quality training examples and understand how AI uses context to make decisions.

Activities

Morning Session: Building Perfect Examples

1. The "Perfect Solution" Template (60 minutes)
• Create a standard format for documenting solutions:
• PROBLEM: [Word problem text]READ: [What we know/What we need to find]BUILD: [Step-by-step with Montessori materials]DRAW: [Visual representation with colors]WRITE: [Mathematical notation/equation]
2. Creating 10 Perfect Examples (45 minutes)
• Working in pairs, students create 10 complete examples
• Cover addition, subtraction, multiplication, division
• Include detailed photos of each Montessori step

Afternoon Session: Understanding Context

3. Context Construction Game (30 minutes)
• Give students partial information about a problem
• Show how additional context changes the solution approach
• Demonstrate: "AI needs to know WHEN to use which method"
4. Organizing Information for AI (45 minutes)
• Sort training examples by:
• Problem type (addition, subtraction, etc.)
• Difficulty level
• Montessori method used
• Common student mistakes

Assessment

Students submit their 10 training examples in the standard format

---

Day 3: Prompt Engineering & Communication with AI

Learning Objectives

Students will learn how to write clear, specific instructions that AI can follow to create accurate visual guides.

Activities

Morning Session: Talking to AI

1. Prompt Engineering Basics (45 minutes)
• Demonstrate the difference between vague and specific prompts
• Show examples:
• Poor: "Make a math picture"
• Good: "Create a step-by-step visual showing 234 + 157 using Montessori stamp game materials with green units, blue tens, and red hundreds in a left-to-right layout"
2. The Perfect Prompt Formula (45 minutes)
• Teach the template:
• CONTEXT: You are creating educational anchor charts for 6th grade Montessori mathPROBLEM: [Specific word problem]METHOD: Use Montessori stamp game with proper color codingFORMAT: Show Read-Build-Draw-Write process in 4 panelsSTYLE: Clear, educational, step-by-step progressionCOLORS: Green=1s, Blue=10s, Red=100s, Green=1000sLAYOUT: Left to right progression showing each step

Afternoon Session: Testing Our Prompts

3. Prompt Testing Lab (60 minutes)
• Students write prompts for their training examples
• Use text-to-image AI to test prompts (teacher facilitated)
• Refine prompts based on results
• Document what works and what doesn't
4. Building the Prompt Library (30 minutes)
• Create a shared document of successful prompts
• Organize by problem type and complexity

Homework

Students refine 5 prompts for their best training examples

---

Day 4: Training Data Quality & AI Accuracy

Learning Objectives

Students will understand how to evaluate AI output and improve training data quality.

Activities

Morning Session: Quality Control

1. Evaluating AI Output (45 minutes)
• Create a rubric for evaluating generated anchor charts:
• Accuracy of math steps
• Correct use of Montessori colors
• Clear visual progression
• Appropriate for 6th grade level
2. Common AI Mistakes (45 minutes)
• Analyze failed AI outputs
• Identify patterns in mistakes
• Develop strategies to prevent these errors in prompts

Afternoon Session: Improving Our System

3. Feedback Loop Creation (60 minutes)
• Students test their refined prompts
• Use the rubric to score results
• Modify prompts based on scores
• Re-test and compare improvements
4. Building Error Prevention (30 minutes)
• Create a "mistake prevention checklist" for prompts
• Add common error fixes to prompt templates

Assessment

Students demonstrate improvement in their AI output quality scores

---

Day 5: System Integration & Advanced Features

Learning Objectives

Students will combine all components into a working system and add advanced features.

Activities

Morning Session: Building the Complete System

1. RAG System Assembly (60 minutes)
• Organize all training data into searchable categories
• Test the system: Input problem → Retrieve similar examples → Generate solution
• Verify the system can find relevant examples for new problems
2. Advanced Prompt Engineering (45 minutes)
• Add dynamic elements to prompts based on problem type
• Include specific instructions for different operations
• Test with complex, multi-step word problems

Afternoon Session: System Testing

3. Beta Testing (60 minutes)
• Students test each other's systems with new word problems
• Document successes and failures
• Provide feedback for improvements
4. Troubleshooting Workshop (30 minutes)
• Address common system failures
• Develop solutions for edge cases
• Create user guides for the system

Assessment

Working system demonstration with peer feedback

---

Day 6: Presentation & Reflection

Learning Objectives

Students will present their AI systems and reflect on the learning process.

Activities

Morning Session: Final Presentations

1. System Demonstrations (90 minutes)
• Each team presents their Montessori Math AI
• Live demonstration with audience-provided word problems
• Explanation of their RAG system and prompt engineering

Afternoon Session: Reflection & Future Planning

2. Project Reflection (45 minutes)
• What worked well?
• What was challenging?
• How could the system be improved?
• What did you learn about AI and teaching?
3. Future Applications (45 minutes)
• Brainstorm other subjects where this approach could work
• Discuss how AI might change education
• Plan next steps for improving their systems

---

Recommended AI Tools & Technical Specifications

Best Text-to-Image Models for This Project

1. DALL-E 3 (via ChatGPT Plus)
• Excellent at following detailed instructions
• Good with educational content
• Handles text within images well
2. Midjourney
• High-quality educational illustrations
• Good consistency across image sets
• Effective with specific style requirements
3. Stable Diffusion (via Hugging Face)
• Free to use
• Can be fine-tuned with custom training data
• Good for classroom use

Optimal Prompt Engineering Template

EDUCATIONAL ANCHOR CHART PROMPT:

 

Context: Create a step-by-step educational anchor chart for 6th grade Montessori mathematics

 

Problem: [INSERT SPECIFIC WORD PROBLEM]

 

Visual Requirements:

- 4-panel layout showing Read-Build-Draw-Write progression

- Panel 1: Problem text with key information highlighted

- Panel 2: Montessori materials arranged to show the solution (Green=1s, Blue=10s, Red=100s, Green=1000s)

- Panel 3: Mathematical drawing/diagram of the solution process

- Panel 4: Final written equation and answer

 

Style: Clean, educational, child-friendly, clear fonts, bright but not overwhelming colors

 

Layout: Horizontal progression from left to right, each panel clearly labeled, consistent spacing

 

Specific Instructions:

- Use authentic Montessori stamp game appearance

- Show regrouping/exchanging process if needed

- Include arrows showing the flow between steps

- Make text large enough for classroom display

- Ensure mathematical accuracy in all representations

RAG System Structure

Database Organization:

• Problem Types: Addition, Subtraction, Multiplication, Division
• Difficulty Levels: Basic, Intermediate, Advanced
• Montessori Methods: Stamp Game, Bead Chains, Place Value Charts
• Common Errors: Typical mistakes and corrections
• Visual Examples: High-quality anchor chart examples

Context Construction Process:

1. Analyze input word problem
2. Identify operation type and complexity
3. Retrieve similar solved examples
4. Extract successful prompt patterns
5. Generate contextual prompt for new problem
6. Include error prevention based on common mistakes

This comprehensive lesson plan provides students with hands-on experience in AI development while deepening their understanding of Montessori mathematics methods.