How do I use this pack?
The Teacher Pack supplements the main page in Teacher mode with everything you need to print. It's designed for one-time printing and multi-year reuse.
Recommended flow
- Preparation: First read through the main page in Teacher mode. There you'll find per chapter: extended learning objectives, timing plan, discussion script, quiz solutions, anticipated student questions.
- Print materials: Print the worksheets below (ideally class set minus 1, keep one spare). Print the parent letter in the number of your class.
- In class: Follow the timing plan in Teacher mode. Worksheets serve as activities, not tests.
- Assessment: Optional class test at the end of the unit. Use the included rubric for grading.
What's included?
- 6 worksheets — one per chapter (2 through 7). Answer key right below each.
- Class test — 28 points across all chapters, with grading rubric.
- 18 homework tasks — 3 tasks per chapter in 3 difficulty tiers.
- Parent letter — editable template with placeholders.
- Curriculum overview — mapping to educational standards (Austrian/German/Swiss).
Printing tips
- Press Ctrl+P (Windows) or ⌘+P (Mac). In the print dialog you can also choose "Save as PDF".
- In print, navigation and background colors are hidden — the result is black-and-white ready.
- Each worksheet starts on a new page. You can select individual pages in the print dialog.
Worksheets per chapter
Each worksheet is suitable for about 15–20 minutes of solo or pair work. The answer key is in a sage-tinted box right below — cut it off before making the class set, or print 2-sided (student side front, answer key back — don't give to student).
Worksheet 1 — What is AI?
1. Definition (2 pts)
Define "Artificial Intelligence" in no more than 2 sentences in your own words:
2. Finding examples (3 pts)
Name 3 concrete AI applications from your everyday life and describe them briefly:
3. Distinguishing (3 pts)
Explain the difference between a normal computer program (e.g. a calculator) and an AI:
4. Explain a term (2 pts)
What does "Machine Learning" mean? Explain with an example:
5. Reflection (2 pts)
Which one concern about AI matters most to you personally? Argue in 1–2 sentences:
Total: ___ / 12
🔑 Answer Key — Worksheet 1
1. AI is a computer program that learns patterns from data and uses them to solve tasks that would normally require human intelligence (e.g. write text, recognise images).
2. Accepted: voice assistants (Alexa/Siri), navigation routing (Google Maps), photo search, translation app, Spotify recommendations, spam filter, Snapchat filters, ChatGPT, car image recognition.
3. Programmed vs. trained: A calculator has fixed rules ("if 2+3, then 5"). An AI has no rules — it learns through examples. Strong answer: explains with the term "training".
4. Learning from data/examples without explicit programming. Example: An AI sees 1,000,000 cat images and learns by itself what makes a cat.
5. Any well-argued answer is accepted. Possible themes: privacy, fake news, jobs, manipulation, algorithmic bias, lack of explainability.
Worksheet 2 — How does an AI learn?
1. Training phases (3 pts)
Name the three training phases of a modern language AI:
2. Quantity comparison (2 pts)
A child can recognise a cat after 5 examples. Why does an AI need millions?
3. Case study (3 pts)
Imagine: An AI was trained only on images of white cats. What problems arise?
4. Match the terms (4 pts)
Match the terms to their definitions (draw lines):
| a) Parameter | → Measure of how wrong the AI still is |
| b) Loss | → AI memorises examples instead of generalising |
| c) RLHF | → Humans rate AI answers; AI learns what's polite |
| d) Overfitting | → "Tuning knobs" in the network; GPT-4 has ~1.7 trillion |
5. Own evaluation (2 pts)
Do you think an AI really "learns"? Or "just" does statistics? Argue:
Total: ___ / 14
🔑 Answer Key — Worksheet 2
1. Pre-Training (with huge datasets) · Fine-Tuning (refinement with curated examples) · RLHF (learning from human feedback).
2. AI combines examples purely statistically in pixel values. A child uses prior knowledge ("four legs, fur, tail = animal") and combines it with the new example. AIs start with no prior knowledge.
3. It might not or poorly recognise black, striped, or spotted cats. It might even surprisingly classify white dogs as "cat". Generally: training-data skew = bias.
4. a→Parameter; b→Loss; c→RLHF; d→Overfitting
5. Pro learning: adapts behavior based on experience. Contra: no consciousness, no understanding, can't learn new examples after training. Any well-argued position accepted.
Worksheet 3 — AI in everyday life
1. AI or not AI? (4 pts)
Tick which of these applications likely use AI:
- ☐ Google Maps (routing)
- ☐ Calculator app
- ☐ Email spam filter
- ☐ Alarm clock / time display
- ☐ Spotify recommendations
- ☐ Snapchat filters (dog ears)
- ☐ Bluetooth pairing
- ☐ Translation app (DeepL / Google Translate)
2. Data question (3 pts)
Pick an AI from your daily life. What data does it need to function?
My AI:
Data it needs:
3. Business model (3 pts)
Spotify gives you personalised recommendations — for free. How does Spotify still make money?
4. Reflection (3 pts)
Which of these AIs would you miss most? Which wouldn't matter to you? Argue:
Total: ___ / 13
🔑 Answer Key — Worksheet 3
1. Uses AI: Google Maps, spam filter, Spotify, Snapchat filters, translation app. NOT AI: calculator, alarm clock, Bluetooth (classical algorithms without learning).
2. Accepted: precise listing. Examples: Spotify → listening habits, skip behavior, time of day, mood; Google Maps → GPS tracks from millions of users, traffic data; Snapchat filter → facial landmarks from thousands of training selfies.
3. Ads (heard in the free version) + Premium subscriptions (15 million pay monthly). Better recommendations = longer use = more ad revenue / higher subscription willingness.
4. Accepted: well-argued answers. Common: nav apps, translation, photo search are missed. Ad recommendations wouldn't matter to some.
Worksheet 4 — Writing prompts
1. The five building blocks (2.5 pts)
A good prompt has five building blocks. Which are they? (0.5 pts per block)
2. Weak prompt → Good prompt (4 pts)
Improve the following weak prompt to a precise, structured prompt:
Weak: "Explain the French Revolution to me"
Your improved prompt:
3. Own use case (3 pts)
Imagine you have to write an essay about your favorite sport. Write a structured prompt that helps you get a good start (the AI should not write the whole essay):
4. Ethics discussion (3 pts)
Should having AI write homework count as cheating? Argue your opinion in 3–4 sentences:
Total: ___ / 12.5
🔑 Answer Key — Worksheet 4
1. Role · Context · Task · Format · Examples (Few-Shot)
2. Example: "You are a history teacher for grade 8. Explain the French Revolution chronologically in 5 steps. Format: numbered list, max 2 sentences each. Focus: political causes and consequences for Europe."
3. Accepted: all 5 blocks used, with clear scoping "intro/brainstorming" instead of "full essay". Example: "You are a sports journalist. I'm writing an essay about football for grade 9. Give me 5 thought-provoking opening questions — please no finished texts."
4. Any well-argued position accepted. Full points: nuanced argumentation, own position, counter-argument mentioned. Middle-ground "AI as brainstorming tool OK, AI as complete writer problematic" is plausible.
Worksheet 5 — What AI can't do
1. Define terms (4 pts)
Briefly explain what these terms mean:
Hallucination:
Knowledge Cutoff:
Bias:
Deepfake:
2. Analyse case study (3 pts)
In 2023, New York lawyer Steven Schwartz filed a court brief with 6 ChatGPT-invented rulings. What should he have done differently?
3. Dangerous areas (3 pts)
Name 3 areas where AI hallucinations can be particularly dangerous:
4. Own fact-check (3 pts)
Ask an AI a specific question about your hometown or favorite team. Describe: What was the answer? Is it correct?
Question:
AI answer:
Verification (source):
Total: ___ / 13
🔑 Answer Key — Worksheet 5
1. Hallucination: plausible-sounding but invented AI statement. Knowledge Cutoff: cutoff date of training data, events after that unknown. Bias: systematic distortion from unbalanced training data. Deepfake: AI-generated fake photo/audio/video.
2. At least 2 of: verify rulings in legal database · cross-check with another AI · use AI only for brainstorming, not as source · have colleagues/interns cross-verify.
3. Accepted: medicine (dosage), law, history/homework, news, finance, life advice. Also accepted with justification: job applications, legal writing, scientific texts.
4. Accepted: documented attempt with concrete, verifiable question. Full points: insight documented (e.g. "the AI got my club's founding year wrong").
Worksheet 6 — Responsibility
1. Four rules (4 pts)
Write 4 rules for yourself about using AI in daily life or at school:
2. Privacy (3 pts)
Which personal information would you NEVER enter into an AI? Name 3 examples and argue:
3. Class rule (4 pts)
Imagine your class decides on its own AI rule. What should be allowed, what not? Write your suggestions as 2 "allowed" and 2 "not allowed" rules:
Allowed:
Not allowed:
4. Future essay (3 pts)
How will AI change your life in 10 years? Write in 4–6 sentences what will improve AND what worries you:
Total: ___ / 14
🔑 Answer Key — Worksheet 6
1. Accepted: any rules showing responsibility. Examples: "Verify sources" · "Never personal data" · "AI as help, not replacement" · "Document own contribution in homework" · "Question critically". Full points: 4 distinct, practical rules.
2. At least 3 of: full name + address · phone number · school/class email · photo ID (selfies) · family info · passwords/codes · bank data · medical records. Reasoning: gives AI provider (often abroad) storage rights — possibly non-removable.
3. Accepted: any consistent, ethically argued solution. Full-point example: Allowed: "AI as brainstorming for topic ideas", "AI for understanding theory". Not allowed: "AI writes whole essays", "AI gives personal answers in job interviews".
4. Accepted: nuanced view (advantages AND concerns). Full points: concrete examples (e.g. "doctors use AI for faster diagnosis but patients may lose trust") instead of generalities.
Class Test — End-of-Unit Exam
Duration: 45 min · Points: 28 · Grade: per rubric below
The test does not fit inside a single 90-min double lesson — recommended as a separate lesson at the end of the weekly module (3 × 50 min) or as the final session of the 6-week project.
Part A: Multiple Choice 1 pt each · 6 pts
1. Which of the following statements best describes AI?
- A computer that thinks like a human
- A collection of real nerve cells
- A program that learns patterns from examples
- A very fast calculator
2. Which of these applications uses NO AI?
- Photo search on smartphone
- Spam filter
- Translation app
- Bluetooth speaker pairing
3. What is an AI hallucination?
- When the AI sees colors
- When the AI sounds plausible but states false things
- When the AI dreams
- When the AI gives no answer
4. Which prompt component is most often missing?
- Role
- Context / audience
- Task
- Examples
5. RLHF stands for…
- Random Learning Heuristic Function
- Reinforcement Learning from Human Feedback
- Recursive Loss Hash Function
- Rapid Lookup Hierarchical Format
6. What does "Knowledge Cutoff" mean?
- The point at which the AI's knowledge ends
- The dividing line between knowing and guessing
- The AI's battery usage
- The maximum number of tokens per request
Part B: Short Answer 2 pts each · 8 pts
7. Explain Machine Learning in 1–2 sentences:
8. Name the three training phases of a modern language AI:
9. What does "Bias" mean in an AI? Give an example:
10. Why can't an AI know yesterday's events?
Part C: Application 8 pts
11. Improve the following prompt to a structured prompt with all 5 building blocks (4 pts):
Weak prompt: "Write something about climate change"
12. An AI tells you: "Wolfgang Amadeus Mozart was born in 1789 in Berlin." What do you do? Argue your next step (4 pts):
Part D: Reflection 6 pts
13. Discuss in 5–8 sentences: "Should having AI write homework count as cheating?" Argue both for AND against, then state your own position.
🔑 Answer Key — Class Test
Part A: 1c · 2d · 3b · 4b · 5b · 6a
Part B:
7. ML is a subfield of AI where computers learn from data instead of being explicitly programmed.
8. Pre-Training, Fine-Tuning, RLHF.
9. Bias = systematic distortion. Example: AI trained mostly on images of white men → recognises other skin tones/genders less accurately.
10. Because the training data has a cutoff (Knowledge Cutoff). Everything after is unknown to the AI.
Part C:
11. Full answer contains all 5 blocks. Example: "You are a climate researcher (Role). I'm a grade 9 student writing a research paper (Context). Explain the 3 main causes of climate change (Task). Format: numbered list, 2 sentences each (Format). Use examples like CO₂ emissions and methane (Examples)." Points: 0.8 per block; synthesis (clearly written) 0 or 1 pt.
12. Ideal answer recognises this is a hallucination (Mozart was born 1756 in Salzburg). Next step: verify sources (Wikipedia, encyclopedia, ask another AI / search engine). Bonus: explanation WHY the AI doesn't know / invents this.
Part D:
13. Full points: nuanced discussion with clear pro/con arguments and own justified position. Middle-ground answers especially valued. Accepted: AI as brainstorming OK, AI as writer problematic · Analogy to calculator in math · Comparison to Wikipedia in homework · Pro: saves time for own reflection · Contra: no own effort, learning doesn't happen.
Grading Rubric
| Points | Grade (US) | Grade (UK) | Notes |
|---|---|---|---|
| 25 – 28 | A | Distinction | Full understanding, independent reflection, precise terminology. |
| 22 – 24 | B | Merit | Solid knowledge, minor gaps, reflection present. |
| 17 – 21 | C | Pass | Basics understood, reflection surface-level. |
| 13 – 16 | D | Borderline | Key terms known, many gaps in application. |
| 0 – 12 | F | Fail | Fundamentals not understood — tutoring recommended. |
Weighting Recommendation
- Knowledge (Parts A + B): 14 pts — clear right/wrong line
- Application (Part C): 8 pts — room for partial credit per task
- Reflection (Part D): 6 pts — grade depth of argumentation, not the position
Homework Collection — 3 Difficulty Tiers
Per chapter, 3 tasks: Easy Medium Challenging. Answer hints are in collapsible details below each (screen-collapsible, always open in print).
Chapter 2 — What is AI?
Find 5 AI applications you used on your smartphone today. Note them with short descriptions.
🔑 Answer hint
Accepted: autocorrect, photo gallery search, voice assistant, recommendations (Spotify, YouTube, TikTok), translation, map routing, spam filter.
Make a table: 5 applications × columns "uses AI?" / "which data?" / "who's behind it?". Research where needed.
🔑 Answer hint
Example entry: "Spotify recommendations | Yes | Listening habits, skip behavior, time of day | Spotify Sweden". Full marks: all 3 columns filled, research source cited.
Write a letter to your grandmother/grandfather (or an 80-year-old person) explaining AI in max one A4 page. Avoid technical jargon.
🔑 Answer hint
Full marks: no technical terms like "algorithm" or "neural network" without explanation, with comparison to something everyday (example: "like a small child who sees many pictures and then recognises cats everywhere"), personally addressed.
Chapter 3 — How does AI learn?
Name the 3 training phases and describe each in one sentence.
🔑 Answer hint
Pre-Training (learns language from huge text corpus) · Fine-Tuning (refined with examples of how to answer) · RLHF (humans rate answers, AI learns what's polite/helpful).
Try out Google's Teachable Machine. Train your own AI for 2 categories (e.g. "open hand" / "closed hand") with 30 images each. Describe in 4–6 sentences what you learned.
🔑 Answer hint
Accepted: documented experiment with clear results. Insights: "more examples = better", "AI recognises poorly with different background", "with only 30 examples it's already surprisingly good". Bonus: reflection on bias from too little image variety.
Research: How much did training GPT-4 cost (an estimate is fine)? Compare with the annual budget of a school of your choice. Write a brief analysis (1 A4 page).
🔑 Answer hint
GPT-4 training: estimated USD 100 million (OpenAI officially vague). Comparison: average secondary school annual budget ~USD 2 million. → AI training = 50 school years. Bonus: reflection on energy consumption + CO₂.
Chapter 4 — AI in everyday life
Find a new AI application not shown on the page. Describe it in 3 sentences.
🔑 Answer hint
Accepted: e.g. bank credit checking, automatic YouTube captions, Apple Music beats recommendations, Skype real-time translation, airport face recognition.
Pick an AI application. Imagine it doesn't work for a week. What would you miss? What would be a good alternative?
🔑 Answer hint
Accepted: concrete daily-life examples. Spotify without recommendations → manual browsing (pro: conscious choice, contra: no discoveries). Bonus: reflection on dependence.
Interview an adult in your circle: 3 questions about AI at work. Write the interview as a mini reportage (max 400 words).
🔑 Answer hint
Accepted: documented interview with clear Q+A + own analysis. Full marks: 3 different thematic angles touched (e.g. "facilitates", "threatens", "changes").
Chapter 5 — Talking to AI
Write a good prompt for: "I need help with a presentation on photosynthesis." Don't forget the 5 building blocks.
🔑 Answer hint
Example: "You are a biology teacher for grade 8. Explain photosynthesis in 5 steps. Format: numbered list, 2 sentences each. Example: sunlight hits leaf → chlorophyll converts to chemical energy."
With an AI of your choice, try 3 differently-worded prompts on the same topic. Compare the answers and write down your observations.
🔑 Answer hint
Accepted: documented experiment with concrete comparison. Insight: vague prompt → vague answer. Bonus: insight into "what made the difference?".
Write a prompt that gets the AI to invent a mini-play with 3 characters about a school topic. Try the prompt and iterate.
🔑 Answer hint
Accepted: documented iteration process. Full marks: at least 3 iterations with clear improvement per step. Bonus: play has recognisable characters and dialogue.
Chapter 6 — Limits of AI
Note 3 areas where AI hallucinations are particularly dangerous. Briefly justify each.
🔑 Answer hint
Accepted: medicine (wrong diagnosis kills) · law (wrong rulings cited) · news (fake news amplified) · finance · science.
Fact-check: Pick a specific historical person or event from your history lessons. Ask two different AIs (e.g. ChatGPT + Gemini) for details. Compare with Wikipedia and document discrepancies.
🔑 Answer hint
Accepted: documented comparison. Bonus: concrete hallucination identified. Insight: AIs disagree, Wikipedia is not 100% either — critical thinking is always needed.
Discussion essay (1 A4 page): "What happens when a teacher grades student work with an AI?" Argue for and against.
🔑 Answer hint
Accepted: balanced argumentation. Pro: consistency, time-saving. Contra: AI has bias, no understanding of individual situation, errors not explainable. Own position with justification.
Chapter 7 — Using AI responsibly
Write your personal "4 rules for using AI" on a piece of paper (big enough to display).
🔑 Answer hint
Accepted: any rules showing responsibility. Examples: "Check sources" · "Never personal data" · "AI as help, not replacement" · "Question critically". Bonus: poster-ready design.
Design a poster campaign (1 A4 or digital) on "AI in school life — what's OK, what's not?". Target: 7th graders.
🔑 Answer hint
Accepted: clearly structured with examples. Full marks: age-appropriate (no jargon), visually appealing, unambiguous.
Write a letter to a future student 30 years from now. How was the world with AI in 2025? What went well? What went badly? (max 500 words)
🔑 Answer hint
Accepted: well-formulated future perspective. Full marks: concrete observations from 2025 (e.g. ChatGPT in schools, deepfakes), nuanced view on positive AND negative developments.
Parent Letter Template
You can adapt this template to your school and class. Replace the [Placeholders in grey] with your details and print for your class.
[Your School]
[Address]
[Date]
To the parents of grade [X]
Subject: Teaching unit "Understanding Artificial Intelligence"
Dear parents,
Over the coming [weeks / double lesson], your child's class will be working intensively on the topic of Artificial Intelligence. AI is now ubiquitous in our daily lives — from photo apps to voice assistants to tools like ChatGPT. We want to empower your children to use AI competently and responsibly.
What your child will learn:
- What is AI actually — and what is it not?
- How does an AI learn? What are training data, what are hallucinations?
- Where do we encounter AI in everyday life — visibly and invisibly?
- How do you talk to an AI? (Prompt engineering)
- What are the limits of AI? When can't you trust it?
- How do I use AI responsibly and protect my data?
Materials and source: We use the freely available teaching platform ki-verstehen.webhoch.com/en/, provided by Austrian Webagentur Hochmeir e.U. under free license (CC BY 4.0). Content is age-appropriate for grade [7/8/9].
How you can support at home:
- Talk with your child about AI tools you use yourself (e.g. voice assistants, translation apps).
- For homework: please discuss our class rule: "AI as research helper OK, AI writing complete assignments NOT OK."
- Help your child check AI answers critically (sources!).
- Protect personal data: never enter names, addresses, passwords into AI tools.
Important notes:
- AI tools like ChatGPT often have an age limit (typically 13+ or 16+). Please check the terms of use.
- In class we use AI exclusively [supervised / via school account / as demo on projector], without entering student personal data.
- For questions or concerns: [Your email address]
We're glad your child is acquiring this important future skill — and value your support along the way.
Kind regards,
[Your name]
[Title / Class teacher]
Curriculum Overview
This unit covers key learning areas across multiple curricula. Adaptable for Austrian, German, Swiss, UK, US, and Australian standards.
Curriculum mapping (Austria: AHS Lower Secondary, similar to Germany Sek I, Switzerland Lehrplan 21, UK KS3, US Middle School)
| Subject | Standard / Competency | Chapters |
|---|---|---|
| Computer Science / Digital Literacy | Explain how digital systems work; identify AI applications | 1–4 |
| English / Language | Argue, discuss, write texts with/without AI assistance | 5–7 |
| Ethics / Religion | Responsibility in a technological world | 6–7 |
| Mathematics | Probability as the basis of machine learning | 3 |
| History / Civics | Industrial Revolution → Automation → AI | 2, 7 |
Competency matrix (Bloom's Taxonomy)
| Level | Name | Example from this course |
|---|---|---|
| K1 | Knowledge | Name terms (AI, ML, algorithm, RLHF, hallucination…) |
| K2 | Comprehension | Explain difference "programmed vs. trained" |
| K3 | Application | Formulate structured prompt, use cat trainer |
| K4 | Analysis | Recognise hallucination in live test, Schwartz case study |
| K5 | Evaluation | Develop own position on "AI = cheating?" |
| K6 | Creation | Train own AI with Teachable Machine (Homework 3.2) |
Time allocation
| Format | Distribution | Recommended for |
|---|---|---|
| Double lesson (90 min) | Intensive workshop, all 7 chapters in excerpts, one worksheet | Project day, substitute lesson, taster course |
| Weekly module (3 × 50 min) | Day 1: Ch. 1–3; Day 2: Ch. 4–5; Day 3: Ch. 6–7 + class test | Standard teaching over one week |
| 6-week project (6 × 90 min) | One double lesson per chapter, own research projects | Depth, project work, gifted student program |
| Project week (5 × 4h) | Days 1–4: chapter depth with own research · Day 5: presentations + test | Themed week, IT summer camp |
Prerequisites & follow-up
- Helpful prerequisites: Basic internet / browser / smartphone knowledge. Familiarity with the term "app".
- Follow-up topics: Programming intro (Scratch, Python), media literacy, data protection (GDPR), ethics discussions.
- Extracurricular: Reference to Teachable Machine (browser tool), Elements of AI (free online course).