Teacher's Guide · CBSE Class 6, 7 & 8 · Part 2 — Artificial Intelligence
👩🏫 The AI Lab — Teacher's Guide
A simple, ready-to-read explanation of every activity in the AI Lab — what it teaches,
what students do on screen, exactly what to say, the key words, questions to ask, and the common mistake to clear up.
How every activity works (the same everywhere):
- Open the page, read the title and the short instruction line under it.
- The big playground is hands-on — students click, type, drag sliders, or press buttons and watch what happens.
- A blue "🤖 In a real AI…" box connects the play to the real idea.
- Every page ends with a scored quiz (Practice 🎯) whose questions come straight from that chapter's exercises in the handbook.
- It all runs in a browser — online at the links below, or offline from the downloaded ZIP (no internet needed).
Suggested 1-period flow: 2 min you demo on the board → 8–10 min students play → 5 min the quiz → 3 min discuss the "Ask the class" questions.
🗺️ Which activity for which chapter
Use this to match a playground to the lesson you're teaching.
| Class 6 chapter | Activities to use |
|---|
| Ch 1 — Introduction to AI & Everyday Examples |
What is AI? · AI Around Us · Supervised (Classification) · Unsupervised (Clustering) · Reinforcement Learning |
| Ch 2 — Basic Data Concepts | Basic Data Concepts |
| Ch 3 — Pattern Recognition & Decision Making | Patterns & Decisions |
| Ch 4 — Ethics & Digital Responsibility | Ethics & Digital Safety |
| Class 7 chapter | Activities to use |
|---|
| Ch 1 — AI Domains & Applications |
Computer Vision · Tokenization · Cleaning words · Word numbers · Guess the next word · Classification · Regression · Clustering · Datasets |
| Ch 2 — AI in Industries | AI Around Us |
| Ch 3 — Data Visualisation & Analysis | Data Visualisation |
| Ch 4 — Ethics & AI Bias Awareness | Bias & Fairness · Ethics & Responsible AI |
| Class 8 chapter | Activities to use |
|---|
| Ch 1 — AI Project Lifecycle | AI Project Lifecycle · Datasets · Training vs Answering |
| Ch 2 — Deeper Dive into AI Applications | AI Around Us · Computer Vision · Classification (no-code idea) |
| Ch 3 — Data & Fairness in AI | Bias & Fairness · Classification |
| Ch 4 — Ethics & Responsible AI | Ethics & Responsible AI |
📋 The activities at a glance
🎒 Class 6 — foundations
🤖Class 6 · What is AI? — intelligence, history & the Turing Test
Class 6 · Ch 1
Big ideaWhat intelligence means, the story of AI, and how to tell real AI from plain automation.
What students doFour games: 1) Which kind of intelligence? (🤝 Interpersonal / 🌿 Naturalistic / 🧠 Intrapersonal) · 2) The story of AI — press ▶ Next milestone through automatons → the Turing Test → 1956 birth → AI Winter → modern AI · 3) The Turing Test — tap 🧑 Human or 🤖 Machine for each reply · 4) Automation or AI? — tap ⚙️ Automation or 🤖 AI.
AI = computing + intelligence. People are clever in different ways. The dream of thinking machines is old — from water-clock automatons, to Alan Turing asking "Can machines think?", to 1956 when the field got its name. But a machine is only AI if it learns from data; if it just follows fixed rules, it's automation.
Key wordsintelligence · Turing Test · automation · machine learning (supervised, unsupervised, reinforcement).
Who introduced the Turing Test? · In which year was AI "born"? · Is a microwave AI or automation?
Not every smart-looking machine is AI — automation follows fixed rules and never learns.
🗂️Class 6 · Basic Data Concepts — types, tables & charts
Class 6 · Ch 2
Big ideaData comes in five types, and we make sense of it with tables, bar charts and pictograms.
What students doPlayground 1 — Sort the data: drop each item into 🔢 Numerical / 📝 Text / 🖼️ Image / 🎬 Video / 🔊 Sound. Playground 2 — Bar chart: a lemonade-sales chart with −/+ buttons; the best-seller is highlighted. Playground 3 — Pictogram builder: trees planted per class, with a switchable key 🌳 = 5 / 🌳 = 10.
Data is just raw facts. It can be numbers, text, images, video or sound. Once we organise it and draw it, patterns pop out — a bar chart compares amounts, and a pictogram counts using a picture plus a key, like 🌳 = 5 trees.
Key wordsdata types · table · bar chart · pictogram · key.
Is CCTV footage video or image data? · Which lemonade sold the most? · In the pictogram, what does one 🌳 stand for?
A pictogram needs a key (🌳 = 5) — without it you can't tell the real numbers.
🔁Class 6 · Patterns & Decisions — spot it, predict it, decide
Class 6 · Ch 3
Big ideaSpot patterns to predict what comes next, make observations from data, then make a good decision.
What students doPlayground 1 — Continue the pattern: pick the next shape, number or colour. Playground 2 — What do you observe?: read a small bar chart and answer. Playground 3 — Decide from what you observe: choose the sensible action for each everyday situation.
A pattern is something that repeats, so it lets us predict what comes next. With data we first make an observation (what do we see?), then draw a conclusion, then make a decision. Finding patterns in data is exactly what machine learning does too.
Key wordspattern · predict · observation · conclusion · decision.
What comes next: 2, 4, 6, 8…? · Which day had the highest attendance? · You scored low marks — what's the smart decision?
A conclusion must come from the data you actually see — not from a guess.
🔐Class 6 · Ethics & Digital Safety — stay safe online
Class 6 · Ch 4
Big ideaUse the internet safely — spot online tricks, build strong passwords, and mind your digital footprint.
What students doGame 1 — Spot the threat: label each situation 🎣 Phishing / 📨 Spam / 💻 Hacking / 📋 Plagiarism / ✅ Safe. Game 2 — Build a strong password: type a password and watch a live meter check length, capital, small letter, number and symbol. Game 3 — Active or passive footprint?: tap ✋ Active (you did it on purpose) or 👻 Passive (happened without you realising).
The internet is useful but you must be careful. Phishing tries to trick you into giving up passwords; copying others' work is plagiarism. Make passwords long, with capitals, numbers and symbols — and never share them. And remember: everything you do online leaves a footprint, sometimes without you even realising.
Key wordsphishing · spam · hacking · plagiarism · strong password · digital footprint (active / passive).
An email says "win a phone, share your password" — what do you do? · What makes a password strong? · Is a website quietly storing cookies an active or passive footprint?
"You won a free prize — click here!" is almost always a trick. Never share passwords, and check before you click.
🗣️ Language — how computers read & write
🔪1. Tokenization
Class 7 · Ch 1 (NLP)
Big ideaComputers can't read words directly. They first chop text into small pieces called tokens and give each a number.
What students doType any sentence (or tap a "try these" like ChatGPT is amazing 🤖 or 2 cats + 3 dogs = 5 pets), then choose how to chop it: By words, By word-pieces (subword), or By letters. Coloured token chips appear, each with an ID number and a total count.
Before a computer can understand a sentence, it cuts it into pieces — like a row of word-cards. Each piece is a "token" and gets its own number, because computers only ever work with numbers.
Key wordstoken = a piece of text · token ID = its number · subword = a chunk smaller than a whole word.
How many tokens are in "I love AI"? · Why might a long word like "unbelievable" get split into pieces? · Do emojis and punctuation get their own tokens?
A token is not always a whole word — long or rare words get split into smaller pieces.
🧹2. Cleaning words
Class 7 · Ch 1 Class 8 · Ch 1 (data cleaning)
Big ideaWe tidy text before an AI uses it — make it lowercase, drop tiny stop-words, and cut words down to their root.
What students doPick a phrase (e.g. studies & runs, leaves & happier) or type their own, then switch steps on/off: a→a lowercase, 🚮 remove stop-words, ✂️ stemming, 📖 lemmatization. Compare Original vs After cleaning.
Computers waste effort on tiny words like "the" and "is", and they get confused that run, runs and running look different. Cleaning removes the clutter and turns words back to their root so the computer sees the real meaning.
Key wordsstop-words = very common words (the, is, a) · stemming = chop to a rough root (studies → studi) · lemmatization = proper dictionary root (studies → study).
What is the root of "running"? · Why remove the word "the"? · What's the difference between stemming and lemmatization?
Stemming can give a non-word (studi); lemmatization gives a real dictionary word (study).
🔢3. Word numbers (embeddings)
Class 7 · Ch 1 (NLP)
Big ideaWords are turned into lists of numbers so that words with similar meaning sit close together on a "meaning map".
What students doClick any word on the meaning map to light up its nearest neighbours. Try "meaning maths": king→queen (make it female) and dog→puppy (make it a baby) — the same "direction" is applied to other words.
AI turns each word into numbers — like coordinates on a map. Words that mean similar things land near each other. Amazingly, directions have meaning too: the step from "king" to "queen" is the same kind of step as "man" to "woman".
Key wordsembedding = a word turned into numbers · neighbour = a nearby, similar word · vector = the list of numbers (an arrow/direction).
Which words sit near "happy"? · What does "king − man + woman" give? · Why is storing meaning as numbers useful?
The numbers aren't random — distance on the map = similarity in meaning.
🔮4. Guess the next word
Class 7 · Ch 1 Class 8 (how LLMs work)
Big ideaA language model writes by repeatedly guessing the most likely next word — a probability game.
What students doPick a starter (The…, I…, Once…, My…, The dog…). The model shows several next-word guesses with probability bars; click one to add it and get fresh guesses. A sentence grows word by word.
A language model doesn't plan the whole sentence. It just keeps asking "what word probably comes next?" Each option has a percentage. Pick words and watch a sentence build — that's exactly what ChatGPT does, only very fast.
Key wordsprediction · probability = how likely · next-token.
After "The dog…", what is the top guess? · Why are there several options, not just one? · Is the highest-probability word always the "correct" one?
It predicts what's likely, not what's true — so it can write silly or wrong sentences.
🤖6. How an LLM learns vs answers (Training vs Inference)
Class 8 · Ch 1 (model training)
Big ideaTwo different jobs: Training = slowly learning from lots of text by adjusting millions of "knobs"; Inference = using what it learned to answer, one token at a time.
What students doTrain one step / ⏩ Auto-train and watch "how often it's right now" climb and the "knobs (weights)" change. Then pick a prompt (The sun rises in the…) and Generate next token → / ⏩ Finish it to watch it answer.
Learning and answering are different. While TRAINING, the model reads tons of text and tweaks tiny knobs until its guesses improve — slow, and done once. While ANSWERING (inference), it just uses those fixed knobs to produce words quickly.
Key wordstraining · weights/knobs · accuracy · inference = answering.
In which phase do the knobs change? · Why does accuracy go up during training? · Is answering fast or slow compared to training?
The model is not learning while it answers you — the learning happened earlier, during training.
👁️ Vision — how computers see
👁️5. Computer Vision
Class 7 · Ch 1 (the Eyes) Class 8 · Ch 2 (image AI)
Big ideaA computer sees a picture as a grid of numbers, and finds shapes by sliding a small filter that detects edges.
What students doDraw on the pixel grid (click squares) or pick a shape (7, A, H, ♥, ＋, 🙂, line). Tap 🔢 show numbers to see it as 0s and 1s. Choose ↕ vertical edges or ↔ horizontal edges, then ▶ Slide the filter and watch the edges light up.
A photo is just a grid of numbers — bright spots are high, dark spots are low. The first thing an image AI does is slide a tiny window across the picture to find edges, where light meets dark. Stack many of these and it can recognise objects and faces.
Key wordspixel = one dot of the grid · filter (kernel) = a small edge-finder · feature = something it finds, like an edge.
What number is a "lit" pixel? · What does the vertical-edge filter light up? · How does finding edges help the computer recognise a "7"?
The computer doesn't really "see" a 7 — it only sees numbers and patterns of edges.
🧠 The ways machines learn
🗂️7. Classification — Supervised Learning
Class 7 · Ch 1 Class 8 · Ch 3
Big ideaLearn from labelled examples, draw a boundary, then sort new things. It needs a teacher (the labels).
What students doPick a problem (🍎 Healthy or junk?, 🐱 Cat or dog?, 🌱 Will it grow?, 📧 Spam or real?, 📚 Pass or fail?). See the labelled dots and the boundary the computer learned. 🎁 Drop a mystery one or click anywhere to test a new point; clear tests to reset.
In supervised learning, a teacher first labels lots of examples — healthy/junk, spam/real. The computer finds a line, a boundary, that best splits the two groups. For anything new, it just checks which side of the line it falls on.
Key wordslabel · supervised learning · boundary · feature (the two axes).
Who provides the labels? · How does it decide a brand-new point? · Name another classification job (hint: a spam filter).
Supervised learning needs labels / a teacher — that's the opposite of clustering.
📈8. Regression — predicting a number
Class 7 · Ch 1
Big ideaClassification's partner — but it predicts a number (not a category) by fitting a trend line.
What students doPick an example (📚 Study hours → exam marks or 🏠 House size → price). ✏️ Move the line to fit the dots (or tap ✨ Best fit). Then 🔮 predict a NEW number — choose an input and read the line's prediction.
Regression predicts a number. We draw the line that best follows the dots. Then for a new input — say 6 hours of study — we read the line to predict the marks. Classification asks "which group?"; regression asks "how much?"
Key wordsregression · trend line / line of best fit · predict.
If you study more hours, does the predicted mark go up or down? · What mark does the line predict at 5 hours? · Is "price" a category or a number?
Regression gives a number; classification gives a category. Don't mix them up.
🪄9. Clustering — Unsupervised Learning
Class 7 · Ch 1 Class 8 (unsupervised)
Big ideaNo labels at all — the computer finds the groups by itself, putting nearby points together (k-means).
What students doPick a real example (🛒 Shoppers, 🍓 Fruits, 🎵 Songs) or 🎲 Random dots. Choose 2 or 3 groups. Press ▶ One step or ⏩ Find groups to watch the centres move and settle; 🔁 Reset or 🎲 New dots to try again.
Sometimes nobody labels the data — we just have a pile of dots. Clustering lets the computer discover the groups on its own: it drops a few "centres", sends each dot to its nearest centre, moves the centres to the middle, and repeats until they stop moving. No teacher needed.
Key wordsclustering · unsupervised learning · centre (centroid) · k = number of groups.
Does clustering use labels? · How does it decide which dots belong together? · What's the biggest difference from classification?
Unsupervised learning has no labels / no teacher — the exact opposite of classification.
🕹️10. Reinforcement Learning
Class 8 (third way to learn)
Big ideaA robot learns by trial and error, collecting rewards (good) and penalties (bad) — like training a pet with treats.
What students doPick a world (🟩 Open field, ⚡ Lava river, 🧱 Wall maze). Press 🎓 Train 1 try or ⏩ Auto-train (100 tries), then 🤖 Watch it play; 🔁 Forget everything to start over. Watch the arrows (its best move per square) and the reward bars climb.
No labelled answers here. The robot tries moves; reaching the 🏆 gives a big reward, falling in the ⚡ lava gives a penalty. Over many tries it remembers what paid off. After training, watch it walk straight to the treasure — it learned the path entirely by itself.
Key wordsreward · penalty · trial and error · policy = the arrows showing the best move per square.
How does the robot learn the path? · Why does it move almost randomly at first? · What everyday thing is this most like? (training a pet)
The three ways AI learns are supervised, unsupervised, and reinforcement — this is the third one.
🗃️ Working with data
🗃️11. Datasets — the food an AI learns from
Class 7 · Ch 1 Class 8 · Ch 1 (data collection)
Big ideaA dataset is the data we teach an AI with. We split it into training / validation / test, and data comes in different types.
What students doPlayground 1 — Split the dataset: choose 60/20/20, 70/15/15 or 80/10/10 and watch 100 items colour into the three groups. Playground 2 — Sort the data: pick an item and drop it in the right bucket — 🔢 Numerical, 📝 Text, 🎬 Multimedia, ⏱️ Time-series, 🗺️ Spatial; ↻ Play again.
An AI learns from a dataset. We split it up: most for TRAINING (learning), some for VALIDATION (checking while it learns), and some kept secret for the TEST — the final exam on data it has never seen. Data also comes in types: numbers, text, images/audio, time-series, and maps.
Key wordsdataset · training / validation / test split · data types · structured vs unstructured.
Why must we keep the test data secret? · What percent did you give to training? · Is a song "numerical" or "multimedia"?
Never test on data the model already trained on — it's like letting a student see the exam answers first.
📊12. Data Visualisation
Class 7 · Ch 3
Big ideaTurning numbers into pictures (charts) helps us spot patterns fast — and data is only trustworthy when it is valid (accurate AND precise).
What students doPlayground 1 — Your report card: change subject marks with −/+ and switch 📊 Bar / 📈 Line / 🥧 Pie; the highest and lowest are highlighted. Playground 2 — Spot the VALID data: pick the dartboard that is accurate and precise; 👀 Reveal all labels. Playground 3 — Sort the data: 🧱 Structured vs 🎒 Unstructured.
The same numbers can become a bar chart (to compare), a line graph (to show change over time), or a pie chart (to show parts of a whole). Pictures help our brain see patterns quickly. And data is only trustworthy when it's valid — both accurate (close to the truth) and precise (consistent and exact).
Key wordsbar / line / pie chart · pattern · trend · precision · accuracy · valid data.
Which chart best shows change over time? · Which dartboard shows valid data? · Is a table of marks structured or unstructured?
A pie chart shows parts of a whole, not change over time. And "precise" is not the same as "accurate".
🛡️ Building & using AI responsibly
⚖️13. Bias & Fairness
Class 7 · Ch 4 Class 8 · Ch 3
Big ideaAI learns only from the data we give it — so unbalanced or unfair data makes an unfair AI.
What students doDemo 1 — Balance the training data: change how many 🏏 vs 🏸 examples there are, press 🧪 Test the AI, and watch its guesses skew toward whichever it saw more of; ⚖️ Snap to 50/50 to fix it. Demo 2 — Name that bias: read a short story and pick the type of bias (data, historical, measurement, algorithmic, human); Next story →.
Machines don't invent bias — they copy it from the data. If we show mostly cricket bats, the AI guesses "bat" for almost everything. Balance the examples and it becomes fair again. That's why we use diverse, balanced data and keep checking the AI.
Key wordsbias · balanced data · data / historical / measurement / algorithmic / human bias · fairness.
Where does AI bias actually come from? · What happened when the data was 90% bats? · How can we make the AI fairer?
"Garbage in, garbage out" — unfair data gives an unfair AI, even when the maths is perfect.
🔄14. AI Project Lifecycle
Class 8 · Ch 1
Big ideaBuilding an AI is a 6-stage journey; and real AI (which learns/adapts) is different from automation (fixed rules).
What students doPlayground 1 — the journey: press ▶ Next stage through the 6 stages (Define problem → Data collection & prep → Model development & training → Evaluation & refinement → Deployment → Monitoring & maintenance); 🔁 Restart. Playground 2 — spot the difference: tap ⚙️ Automation or 🤖 AI for each scenario. Playground 3 — a tiny model: set attendance, study hours, marks and participation (Yes/No), then test Student 4 / Student 5 to predict Above/Below 75%.
Real AI projects follow steps: first define the problem, gather and clean data, build and train the model, test and improve it, deploy it, then keep monitoring it. And remember — a fixed-rule machine like a timed car wash is just automation; it's only AI if it learns and adapts.
Key wordslifecycle · the 6 stages · automation vs AI · accuracy · refinement · deployment · monitoring.
What is the very first stage? · Is a traffic light that changes every 30 seconds AI or automation? · Why must we keep monitoring after deployment?
Not every machine is AI — automation follows fixed rules and never learns from data.
🌍15. AI Around Us (Applications)
Class 7 · Ch 2 Class 8 · Ch 2
Big ideaWhere AI helps in real life — healthcare, environment, transport, education, agriculture, smart homes.
What students doExplore AI around us: tap a sector to see real ways AI helps (with Indian examples). Match the AI to its field: choose the correct field for each application. ⚙️ Automation or 🤖 AI?: decide which one each scenario is; ↻ Play again.
AI is already all around us — spotting diseases in X-rays, watching for plastic in the oceans, timing traffic lights, giving instant quiz feedback, advising farmers, and running smart homes. Explore each area, then test yourself by matching real examples to the right field.
Key wordsdomain / field · application · healthcare / environment / transport / education / agriculture AI.
Name one way AI helps farmers. · Which field is "analysing MRI scans"? · Is soil-moisture-based irrigation AI or automation?
AI usually assists experts, it doesn't fully replace them — e.g. AI helped steady the surgeon's hands; it did not do the surgery alone.
🛡️16. Ethics & Responsible AI
Class 8 · Ch 4 Class 7 · Ch 4 (digital citizenship)
Big ideaUsing AI the right way — protect your privacy, spot misinformation, and keep humans accountable.
What students doGame 1 — App permissions: tap ✅ Allow or 🚫 Deny for each request and learn what an app really needs. Game 2 — Trust it or Check first: tap ✅ Trust it or 🔍 Check first on each message and learn the verify checklist. Game 3 — Who is accountable: decide who is responsible when an AI makes a decision (Human-in-the-Loop). 🔁 Play again.
Powerful AI needs rules. Protect your privacy — only allow an app the permissions it truly needs, and never share OTPs or passwords. Don't believe everything online — check the source before you share. And remember: a human must always stay responsible for important AI decisions.
Key wordsethics · privacy · permissions · misinformation · accountability · Human-in-the-Loop.
Should a torch app get access to your contacts? · What should you do before forwarding a shocking message? · Who is responsible if an AI makes a mistake?
"The computer decided" is never an excuse — the people and organisations who build and use AI stay responsible.