Session 40: Using AI Wisely & Safely
At a glance: pupils learn how chatbots produce answers, write C.L.E.A.R. prompts, fact-check AI answers, recognise deepfake scams, keep private information out of AI tools and use AI honestly. · Time: 90 minutes (adjust the timings if your period is shorter) · Workbook pages: 6
Learning outcomes
By the end of the session, pupils can:
- Explain that a chatbot predicts likely words, so its answers can sound sure but be wrong ("hallucinations").
- Write a clear prompt using C.L.E.A.R., and check an AI answer against a trusted source.
- Keep their personal information out of AI tools, spot a deepfake scam and tell honest use of AI from cheating.
Before the session
- Test the AI Helper Check lab on the projector PC and a pupil PC. It uses a pretend AI with pre-written answers, so it needs no internet and no account.
- Decide whether you will show a real chatbot today. If so, use your own account on the projector PC, check your organisation's and the school's rules, and type the prompts yourself. Pupils under 13 must not use their own accounts, and under-18s need a parent's or guardian's permission (most AI tools set these limits).
- Turn off chat history or memory in your AI tool if you can, and never type pupils' names into it.
- Have a Social Studies textbook (or a printout) ready for the fact-check on Page 3.
- Prepare a short note to parents and guardians explaining the optional supervised home task (Page 5, Mission 2), if you plan to send it home.
Running order
| Time | Activity | Workbook | What the mentor does |
|---|---|---|---|
| 0–10 min | Typing warm-up | – | Pupils open the typing trainer (Typing in the menu, or the warm-up banner on the session page) and do the 10-minute warm-up. Walk round checking posture and home-row fingers. |
| 10–18 min | From Session 39 to chatbots | Page 1 | Recall "AI is a pattern-guesser". Play next-word prediction: say "On Saturday I will go to the…" and let the class shout words. |
| 18–25 min | Age rules and supervision | Page 1 | Explain the club rule: real AI only with a mentor, at the Hub or with a parent. |
| 25–38 min | Prompts and C.L.E.A.R. | Page 2 | Do the "send a cousin to the market" role-play. Pupils rewrite the two vague prompts in pairs. |
| 38–50 min | AI Helper Check lab | Page 5 | Demonstrate the prompt builder on the projector, then pairs do all three missions and fill in the table. |
| 50–61 min | Hallucinations and the 3 checks | Page 3 | Do the fact-check challenge together with the textbook. |
| 61–72 min | Deepfakes, privacy and honesty | Page 4 | Link deepfakes to S.T.O.P. from Session 33. Vote on the Helper-or-Cheater table. |
| 72–80 min | Real AI demo (optional) | Page 5 | Type 2–3 pupils' C.L.E.A.R. prompts into a chatbot on the projector. Fact-check one answer as a class. |
| 80–86 min | Scenarios and review questions | Page 6 | Discuss each scenario aloud. |
| 86–90 min | Self-assessment and sign-off | Page 6 | Sign off using the evidence below. |
Key ideas and how to explain them
- Chatbots predict words: trained on huge amounts of text, they write answers one likely word at a time, like a super-powered phone keyboard. Likely-sounding is not the same as true.
- Hallucination: a confident, made-up answer (wrong dates, invented books or links). Always do the 3 checks: a trusted source, does it make sense, a second source.
- C.L.E.A.R.: Context, Limits, Examples and details, Ask one clear task, Review. The market errand ("buy food" vs "6 eggs and a tin of tomatoes under 50 cedis") makes it land.
- Deepfakes: AI-made fake photos, videos and voices. The defence is the same as for phishing: panic and money requests are red flags, so stop and check another way (call back on a known number).
- Privacy: what you type may be stored and used for training. Treat a chatbot like a public noticeboard: no P.I.I., no passwords, no photos of other people.
- Honesty: AI as a tutor (explain, quiz me) is fine; AI as a ghost-writer for work you hand in is cheating. Follow school rules and say when you used AI.
The 3D lab
AI Helper Check has three missions on a pretend AI phone. On the projector, demonstrate the first: send the vague prompt, then add pieces (the task, who it is for, the format and a tone) and watch the clarity meter and the answer improve. Pairs then check the helper's answers. Some are confidently wrong, and Check the source opens the textbook. Last, they sort what is OK to ask AI and what must never be typed in. Ask pairs which wrong answer surprised them most. It is the same lesson as Page 3: it sounded sure, and it was wrong. No real AI is used, so this is safe for every pupil.
Common mistakes and quick fixes
| Pupils often… | Try this |
|---|---|
| Believe an answer because it is long and polite | Point to the fact-check challenge: neat writing, wrong facts. |
| Write a prompt with no task ("I am a JHS 1 pupil") | Ask "What exactly do you want it to do?" The A in C.L.E.A.R. is the task. |
| Think a chatbot is searching the internet like Google | Some tools can search, but the answer is still written by prediction, so check it. |
| Think deleting a chat deletes it from the company | It may already be stored. Not typing private things is the only safe choice. |
| Feel any use of AI for school is cheating | Show the Helper column: explaining, quizzing and brainstorming are fine when you do the work yourself and are honest. |
When things go wrong
- No internet: the lab and all six pages work offline. Skip the real AI demo, or show screenshots you prepared before.
- Not enough PCs: threes for the lab, rotating the mouse each mission; the others read the answers aloud and check the textbook.
- A PC freezes or the power goes: do the fact-check, the prompt rewrites and the Helper-or-Cheater vote on paper. Read out "AI answers" from the lab missions yourself and let pupils hold up ✅ or ❌ cards.
- The real AI gives an unsuitable or upsetting answer: stop sharing the screen calmly and say "this is exactly why we check AI and why adults supervise". Report it using the tool's feedback button later.
Answer key
Your Turn: Fix These Prompts (Page 2): open task. Look for all five parts, for example: "I am a JHS 1 pupil in Ghana. Explain how plants make food (photosynthesis) in 5 simple steps, using examples from a cocoa farm." And: "I am a JHS pupil from Cape Coast. Write a 100-word paragraph for a school project describing 3 interesting things about my town for visitors." (Pupils should name a town, not their street or house.)
Fact-Check Challenge (Page 3)
- Mistake 1: independence was on 6 March 1957, not 1960. (1960 is when Ghana became a republic, with Nkrumah as its first president.)
- Mistake 2: the capital is Accra, not Kumasi.
- Kwame Nkrumah and Lake Volta are correct. Lake Volta is one of the largest man-made lakes in the world by surface area.
- Trusted sources: a Social Studies textbook, a teacher or mentor, an official
.gov.ghwebsite, an encyclopedia.
Helper or Cheater (Page 4): 1. Helper. 2. Cheater. 3. Helper. 4. Cheater.
Lab table (Page 5): Mission 1: adding the task, who it is for and the format (a tone is optional) made the answer better. Mission 2: 3 of the 6 answers were wrong (the independence date, "a spider is an insect" and "the Sun goes around the Earth"). Mission 3: for example a password or PIN, a home address, a full name with a school, or photos of friends.
Real AI tool (Page 5, Mission 2): open task. Look for a C.L.E.A.R. prompt with no P.I.I., a one-sentence summary, one fact checked in a named source and a sensible follow-up question.
Scenario Decisions (Page 6)
- Checking a chatbot's date in the textbook: Wise.
- Typing full name, school and home address: Unwise (P.I.I.).
- Calling the uncle back on his usual number before doing anything: Wise.
- Sharing a shocking video straight away: Unwise (it could be a deepfake. T.H.I.N.K.: is it true?).
Written Review Questions
- A chatbot writes by predicting words that are likely to come next, from patterns in its training text. It does not check whether something is true, so it can sound sure and still be wrong.
- Context, Limits, Examples and details, Ask, Review.
Support and stretch
- Needs more support: do the lab's prompt builder together first, then rewrite just one vague prompt with a sentence starter ("I am a… Explain… in…"). Read the fact-check paragraph aloud.
- Finished early: write a C.L.E.A.R. prompt that would help them revise for an upcoming test. Or write three tips for parents on using AI safely at home, to share with their family.
Evidence for sign-off
The pupil writes at least one full C.L.E.A.R. prompt, finds both mistakes in the fact-check challenge with the correct facts, completes all three AI Helper Check missions, and gets all four scenarios right. They can explain in their own words why a chatbot can be confidently wrong.
Safety and wellbeing
Pupils must not create AI accounts in the club or use their own: use a mentor's account on the projector, or supervised access at the Hub. Never type pupils' names, photos or other personal details into an AI tool. Deepfakes and voice-clone scams can be frightening; reassure pupils that the S.T.O.P. habit protects them, and that they should tell a trusted adult about any strange call or message. If a pupil mentions being targeted by fake images of themselves, listen calmly, don't promise secrecy, and follow your school's safeguarding procedure. When suggesting the home task, make clear it is optional and only with a parent's or guardian's permission and supervision.