Class 3 · CBSE AI · Strand D — AI Senses
How do you test an AI? An AI's report card — for kids
AI is tested on examples it has never seen before. How we know whether an AI actually learned. For Class 3 kids.
What this concept actually says
- AI performance is measured by testing it on examples it hasn't seen before
- Accuracy is the simplest measure: how often the AI gets the right answer
- Knowing where an AI fails is as important as knowing its overall score
An analogy your child will recognise
School report card
Your report card doesn't just say 'Priya scored 75% overall.' It breaks it down: Maths 90%, English 80%, Hindi 55%. That breakdown tells you — and your teacher — where to focus. An AI's report card works the same way: you need to know which categories it struggles with, not just its overall score.
Cricket batting scorecard
A batsman's average runs per match is useful, but a cricket coach wants more detail: 'How does he perform against fast bowling? Against spin? In the evening?' That breakdown reveals the real strengths and weaknesses. AI evaluation needs the same detailed breakdown.
Common misconceptions to watch for
- High overall accuracy means the AI is good at everything
- Testing the AI on its own training examples is a valid way to measure its performance
Key facts in one breath
- AI performance is measured by running it on a test set — examples it has never seen during training
- Accuracy is the percentage of test examples the AI answered correctly
- A high overall accuracy can hide serious failures in specific categories
- Analysing errors by category helps you improve the AI more effectively
How Dhee Learning teaches this — the 3-stage question loop
Every Dhee Learning session for this concept follows three stages. We share the questions Dhee actually asks, so you can hear what a session sounds like.
Stage 1 — Surface
In your school exams, you get marks for the questions you answered correctly. How do you think we could give an AI a 'mark' — how would we measure whether it learned well?
Rote answer
"You test it and see how many it got right"
Understood
"You give the AI questions it's never seen before — like a surprise test — and check how many it gets right. That percentage is its accuracy score"
Stage 2 — Reasoning
Your cat-dog AI scores 90% accuracy. Brilliant! But then you notice: it got every single dog wrong and every single cat right. Is 90% a good score here? What does this tell you about using just one number to judge an AI?
Follow-up Dhee may use: Can you think of a real situation where it would be really bad if an AI always got one category wrong? (Hint: think about medical tests or safety systems.)
Stage 3 — Application
Here is your spoon-vs-fork AI's report card: spoon photos tested: 10, got right: 9. Fork photos tested: 10, got right: 6. What should you work on next, and why?
Misconception Dhee watches for: Child thinks overall score is the only thing that matters — not looking at performance by category
Related concepts
Want your child to actually understand this?
Dhee turns this concept into a 15-minute spoken session — asking, listening, and probing — so your child builds the idea themselves.
Frequently asked questions
What is your ai's report card — explained for kids? +
AI is tested on examples it has never seen before. How we know whether an AI actually learned. For Class 3 kids.
What's the most common mistake children make about this concept? +
High overall accuracy means the AI is good at everything
How does Dhee Learning teach this in a Class 3 session? +
Dhee opens with a question — for example: "In your school exams, you get marks for the questions you answered correctly. How do you think we could give an AI a 'mark' — how would we measure whether it learned well?" — listens to your child's answer, then probes the reasoning behind it. The session ends when the child can apply the idea to a brand-new situation, not just recall it.