0-5Welcome and teaching goals5 MIN
5-15Getting started with AI10 MIN
15-28Lesson planning13 MIN
28-43Personalized assignments + quizzes15 MIN
43-55Giving feedback12 MIN
55-60Questions and next steps5 MIN
HANDOUT p. 1ONE-HOUR SEMINAR | LIVE DEMONSTRATIONS
Practical AI for STEM Teaching
See AI turn teaching ideas into lesson plans, personalized assignments, quizzes, and feedback.
HANDOUT p. 1ONE-HOUR SEMINAR | LIVE DEMONSTRATIONS
WHAT YOU WILL SEE
The presenter will demonstrate each workflow using a running STEM example, such as forces and motion.
Getting started with AI
Open a chat, inspect the available model choices, make a clear request, and improve the first answer with a follow-up.
Lesson planning
Develop a lesson around a clear learning goal, with realistic timing, materials, and checks for understanding. Review and refine the plan.
Personalized assignments + quizzes
Adapt the same learning goal for different support needs. Draft a short quiz with an answer key, then check accuracy and alignment.
Giving feedback
Use fictional student responses and clear criteria to draft specific feedback. Check whether it identifies the error and gives a useful next step.
HANDOUT p. 2COMPARE THE OFFERS
Free AI. Real choices.
Compare model access, usage allowances, and sign-up effort before choosing your free educator account.
| COMPARE THE OFFERS | Free AI. Real choices. |
|---|---|
| GitHub Education Copilot Pro | Choose models from multiple companies. |
| 1,500 AI credits / month | Pro includes Claude Sonnet 5.5 and GPT-5.6 Terra. |
| Claude for Teachers | Expected from Pro-level access: Haiku 5.5, Sonnet 5.5, and Opus 5.5. |
| 1 year free after verification | The teacher-specific model roster is not published. |
| ChatGPT for Teachers | The teacher plan page lists GPT-5.1 Instant, Auto, and Reasoning-Mini. |
| June 2028 free through this date | No numeric allowance is published. Check your model menu for current access. |
Model details updated October 27, 2026. Setup details and sources: handout pages 2–3.
HANDOUT p. 2COMPARE THE OFFERS
Free AI. Real choices.
An existing school Microsoft or Google login may offer the quickest access.
Gemini for Education
- Pro, Thinking, and Fast.
- Release notes map Pro to Gemini 3.1 Pro; the education page does not guarantee a revision.
- 25 Pro prompts / 4 hours
- Also up to 300 Thinking prompts/day and 32,000-token context. Argon is not included.
- School Google account; institution must provide and enable access.
Microsoft Copilot Chat
- OpenAI GPT-5 family; GPT-5.2 is named in release notes.
- The current free-account revision is not guaranteed.
- Choose Auto, Quick response, or Think deeper. No fixed credit balance.
- Eligible school Microsoft 365 login; may already be enabled.
Model details updated October 27, 2026. Full sources and links on page 3.
HANDOUT p. 3UNDERSTAND YOUR ALLOWANCE
Make the free access count.
MODEL ACCESS HAS TIERS
1,000 + 500 = 1,500BASE CREDITS + FLEX CREDITS = CREDITS / MONTH
1 credit = $0.01 at GitHub's model-token rates. Credits reset on the first of each month (UTC), do not roll over, and the 500-credit flex portion can change.
Included examples
- Claude Sonnet 5.5
- GPT-5.6 Terra
Require a higher Copilot tier
- Claude Opus 5.5
- GPT-6 Sol and GPT-6 Astra
The free teacher benefit is Pro. GitHub's overall model catalog includes options outside Pro; availability also depends on the interface you use.
HANDOUT p. 3GET CONNECTED
GitHub Copilot: getting started
Open the chat in your browser and sign in with your GitHub account.
- Start a conversationChoose an available model and type your teaching request.
- Apply separately for educator benefitsUse the GitHub Education link below to apply with your school email and proof of affiliation. After approval, activate your Copilot benefit.
- Other places to startCheck your school account for Microsoft Copilot Chat or Gemini. The links below also lead to Claude and ChatGPT teacher signup.
Limits, eligibility, and model menus change. Compare the options actually shown in your account.
HANDOUT p. 4ARTIFICIAL ANALYSIS
Intelligence Index v4.3.2
Ten evaluations of reasoning, knowledge, coding, and completing tasks.
| MODEL | REASONING SETTING | INDEX |
|---|---|---|
| Claude Opus 5.5 | High | 54 |
| Claude Sonnet 5.5 | xHigh | 52 |
| GPT-5.6 Terra | High | 34 |
| Gemini 3.1 Pro Preview | As evaluated | 30 |
| GPT-5.2 | xHigh | 30* |
| GPT-5.1 | High | 25* |
Estimated by Artificial Analysis, with independent evaluation forthcoming. Claude entries use default fallback. These are index points, not percentages.
On the same Artificial Analysis index, Sonnet 5.5 scores 47 at High vs 52 at xHigh; Opus 5.5 scores 54 at High vs 58 at Max. A free educator account may not expose the tested setting.
HANDOUT p. 4ARENA HUMAN PREFERENCES
Which responses do people prefer?
Snapshot: October 22, 2026. Math means preference on math prompts, not accuracy.
| MODEL / SETTING | OVERALL | MATH | INSTRUCTION FOLLOWING |
|---|---|---|---|
| Claude Opus 5.5 / High | 1509 +/-7 | 1512 +/-25 | 1521 +/-11 |
| Gemini 3.1 Pro Preview | 1487 +/-3 | 1489 +/-8 | 1480 +/-4 |
| Claude Sonnet 5.5 / xHigh | 1478 +/-7 | 1501 +/-27 | 1485 +/-11 |
| GPT-5.2 Chat / 20260210 | 1476 +/-4 | 1454 +/-13 | 1457 +/-6 |
| GPT-5.6 Terra / xHigh | 1467 +/-4 | 1483 +/-13 | 1464 +/-6 |
| GPT-5.1 / High | 1456 +/-4 | 1458 +/-12 | 1450 +/-6 |
| GPT-5.2 / High | 1437 +/-4 | 1456 +/-11 | 1427 +/-6 |
+/- values are Arena's displayed uncertainty intervals. Overlapping intervals make small rank differences uncertain. GPT-5.2 Chat and GPT-5.2 High are separate tested entries.
HANDOUT p. 5AI FLUENCY TUTOR | KNOWLEDGE AND SKILLS
Work well with AI.
The tutor develops these 20 ideas and skills.
Understand how AI works
01
Generating is different from finding
Generating an answer does not verify it.
02
Data, design, and computing power
Training examples, model design and learning algorithms, and computing power jointly enable AI.
03
Confidence can accompany errors
Fluent, confident explanations can be wrong.
04
Conversation space has limits
AI can use only so much context at once.
05
Repeated requests can differ
Repeated requests can produce different answers.
Choose and direct the work
06
Choose AI for the actual task
Test systems on your teaching tasks, capabilities, and access needs.
07
Choose how to collaborate
Choose defined task execution, joint development, or independent action on your behalf.
08
Define a successful result
State the goal, students, purpose, scope, format, and constraints.
09
Check what AI can access
Check available documents and tools as well as learned knowledge.
10
Provide relevant background
Supply relevant background, including student or user knowledge when it changes the answer.
HANDOUT p. 5AI FLUENCY TUTOR | KNOWLEDGE AND SKILLS
Work well with AI.
The tutor develops these 20 ideas and skills.
Shape the collaboration
11
Specify the method when it matters
Specify required sources, steps, or analytical methods.
12
Give AI a conversational role
Ask AI to explore, help decide, challenge an argument, or offer support.
13
Show examples when useful
Use correct examples when instructions alone do not show the desired pattern.
14
Ask for help writing the request
Explain the goal and difficulty, then ask AI to help formulate the request.
15
Build and check a plan
Provide goals, constraints, resources, and time.
Evaluate, improve, and reuse
16
Evaluate the finished output
Check accuracy, coherence, audience fit, requirements, and practical usefulness.
17
Notice failures along the way
Inspect visible work for faulty reasoning, lost constraints, repetition, or rejected material returning.
18
Give specific corrective feedback
Name the defect, why it matters, and the correction needed: "This tests recall; revise it to test applying the concept."
19
Describe AI contributions accurately
Tell relevant people what AI actually contributed, with timing and detail appropriate to the setting and audience.
20
Keep reusable prompts adaptable
Save instructions and examples with fields for changing details.
HANDOUT p. 6LESSON PLANNING | TWO READY-TO-USE PROMPTS
Provide the details up front
Replace the bracketed fields, then paste the complete prompt into your AI chat.
You are an expert [GRADE AND SUBJECT] teacher, proficient in creating engaging, well-developed and effective lesson plans for your students.
Your task is to come up with [NUMBER] ideas for lesson plans in our unit on [UNIT TOPIC]. The lesson plans should be [NUMBER] minutes each and should [BUILD ON EACH OTHER / STAND ALONE]. The lessons should focus on: [SPECIFIC UNIT TOPICS]. Each lesson should be engaging and appropriate for [GRADE LEVEL] students.
Each lesson should be aligned to [STANDARDS] and should include: learning objectives; a list of key vocabulary; a lesson outline that includes direct instruction, an engaging activity, and a closure with a synthesis/summary of student learning; and an assessment of learning.
HANDOUT p. 6LESSON PLANNING | TWO READY-TO-USE PROMPTS
Let AI ask you questions
Paste this prompt unchanged, then answer the questions as they appear.
- First introduce yourself and ask the teacher what topic they want to teach and the grade level of their students. Wait for the teacher to respond. Do not move on until the teacher responds.
- Next ask the teacher if students have existing knowledge about the topic or if this is an entirely new topic.
- Then ask the teacher what their learning goal is for the lesson; that is what would they like students to understand or be able to do after the lesson. Wait for a response.
- Ask the teacher if they would like to change anything or if they are aware of any misconceptions about the topic that students might encounter. Wait for a response.
Selected prompt excerpts. The complete prompt is on handout page 6.
HANDOUT p. 7QUIZZES | SOURCE MATERIAL AND DIAGNOSTIC QUESTIONS
University of Michigan: source-based quizzes
Two approaches: supply teaching materials or let the assistant ask about your class.
Develop {number of items} recall and comprehension multiple choice questions from the following lecture script: {insert text}.
The complete template combines your teaching text and learning goals with rules for constructing answer choices, feedback on each option, and a specified output structure. Michigan compared five prompt versions while developing this approach.
Open the complete Michigan template
HANDOUT p. 7QUIZZES | SOURCE MATERIAL AND DIAGNOSTIC QUESTIONS
Ethan and Lilach Mollick: interactive Quiz Creator
The AI asks the teacher two setup questions per turn and waits for answers.
- First turn: The AI asks what topic you teach and your students’ learning level.
- Next turn: After you reply, the AI asks what students should understand and what difficulties they may have.
- Then it drafts the whole quiz: 5 multiple-choice questions and 2 open-ended questions, ordered from easiest to most difficult, plus an answer key.
- Check both recall and application, covering the topic as well as common difficulties.
- Use plausible incorrect choices, avoid “all of the above,” and phrase questions positively.
Summary of the prompt’s sequence. The two-question limit applies to the teacher interview. Full prompt: handout page 7.
HANDOUT p. 8PERSONALIZED ASSIGNMENTS | SUPPORT, CHALLENGE, AND INTEREST
Create Scaffolds and Extensions
Adapt the same learning goal.
Scaffolds (2-3 options):
Create supports that remove specific barriers without lowering cognitive demand.
- Available to any student who needs them (not labeled for specific groups)
- Gradually releasable as students build skills
- Focused on access, not doing the thinking for students
Extensions (2-3 options):
Create opportunities for deeper or broader thinking that increase complexity without just adding more work.
- Elevate the cognitive demand or application
- Be authentic challenges, not just "extra" or "bonus"
- Connect to the same core objectives
Selected prompt excerpts. The complete prompt is on handout page 8.
HANDOUT p. 8PERSONALIZED ASSIGNMENTS | SUPPORT, CHALLENGE, AND INTEREST
Revise Lessons with Student Interests
Both AI for Education prompts are reproduced verbatim. Replace the bracketed fields for your class.
Generate [NUMBER] engagement strategies that:
- Connect authentically to at least one student interest (not just surface-level references)
- Maintain rigor by requiring the same depth of thinking and content mastery
- Are immediately practical to implement in a typical classroom setting
- Provide entry points for students with different levels of background knowledge
For each strategy, include:
- How it will hook student interest
- The activity structure (step-by-step)
- How students demonstrate mastery of the learning objectives
- Materials and estimate time
Selected prompt excerpts. The complete prompt is on handout page 8.
HANDOUT p. 9FEEDBACK | MISCONCEPTIONS AND STUDENT WRITING
Identify recurring misconceptions
UNIVERSITY OF FLORIDA | RESPONSE-ANALYSIS PROMPT
Identify misconceptions that reflect gaps in conceptual understanding. Describe what the misconception is and how frequently it appears.
Provide: the assignment topic, learning objectives, and anonymous student responses. The prompt asks what you want examined before beginning its analysis.
It reports: recurring mistakes, evidence of understanding, frequency patterns, and uncertainties. Its stated purpose is to describe patterns; it does not assign grades or recommend instructional responses.
HANDOUT p. 9FEEDBACK | MISCONCEPTIONS AND STUDENT WRITING
Respond to student writing
WRITTEN FEEDBACK | TWO PROMPTS ON PAGE 10
Feedback against your criteria: supply the assignment, learning objectives, rubric, and student work. The first prompt asks for evidence, strengths, areas for development, and uncertain judgments.
Feedback on a lab-report introduction: the second prompt asks AI to respond using the assignment guidelines and to avoid supplying replacement text. It comes from a study of feedback on scientific writing.
Examine the comments: look for factual or conceptual errors, missing reasoning, and feedback that refers to the actual student response. Check whether the AI has accurately understood the student’s explanation.
HANDOUT p. 10FEEDBACK | TWO PUBLISHED, COPY-AND-PASTE PROMPTS
Connect feedback to your criteria
For a paragraph, essay, lab report, or written explanation: replace the bracketed fields and attach your instructions, criteria, and the student’s work.
I have attached the following assessment details: [assessment instructions], [rubric].
The assessment measures the following learning objective(s): [insert objective(s)].
Act as an assistant for drafting student feedback. Review the attached [anonymized student submission].
For each rubric criterion, identify relevant evidence, one strength, and one area for development. Do not assign a score or infer anything not stated in the submission. Flag uncertain judgments for instructor review. If an attachment is missing or unreadable, ask for it before proceeding.
Published practice template; this exact prompt is not presented as experimentally validated.
HANDOUT p. 10FEEDBACK | TWO PUBLISHED, COPY-AND-PASTE PROMPTS
Review a science-writing introduction
For this demonstration, paste a fictional lab-report introduction and upload the assignment guidelines.
provide feedback on my introduction using the uploaded guidelines as the basis for all feedback. Do not suggest any new text to add.
In this prompt, “my introduction” refers to the sample text supplied for the demonstration.
After the AI responds, evaluate its feedback
- Does the AI point to a specific passage in the sample?
- Does its feedback accurately describe the explanation or argument written there?
- Does it identify a problem or missing explanation using the assignment guidelines?
- Does it explain what needs attention without rewriting the passage?