[ CBSE CLASS 8 / ARTIFICIAL INTELLIGENCE / CHAPTER 3 / 2026-27 ]
Class 8 AI: Data and Fairness – Questions and Answers
The Class 8 Data and Fairness chapter shows that an AI is only as fair as its data. Students check datasets for balance, representation, labelling errors and consent, and compare a model’s accuracy across groups to judge whether it is fair.
[ WORKED EXAMPLE ]
What is a representative dataset?
Answer: One that includes all the kinds of people or cases the AI will meet.
- An AI meets many different people and cases in real life.
- Its training data should include them all.
- Otherwise it may work poorly for missing groups.
What this chapter is about
Students learn what makes a dataset representative: it should include all the people or situations the AI will meet.
They spot imbalance (900 cat photos versus 100 dog photos), missing groups, wrong labels and data collected without consent.
They measure fairness by comparing accuracy for each group, not just overall.
[ HANDS-ON ACTIVITY ]
Sort it: fair data or a fairness problem?
Decide whether each dataset is a fair starting point, or has a problem to fix.
5 left · choose an item
Where does “900 cat photos and 100 dog photos for a cat–dog model” belong?
[ PRACTICE MODE / 6 QUESTIONS ]
Try it yourself
No login. Your score is saved only on this device.
What is a representative dataset?
All Data and Fairness questions with answers
Try each one first, then open it to check the answer and the reasoning.
Q1What is a representative dataset?
- One that includes all the kinds of people or cases the AI will meet
- One that is very small
- One with only one kind of example
- One that nobody can see
Answer: A. One that includes all the kinds of people or cases the AI will meet
Hint: Representative means it “represents” the real situation.
- An AI meets many different people and cases in real life.
- Its training data should include them all.
- Otherwise it may work poorly for missing groups.
Skill: Abstraction · Level: easy · Report an error
Q2A dataset has 900 cat photos and 100 dog photos. What is the likely problem?
- The model may lean towards predicting “cat” and do worse on dogs
- The model will be perfect
- Dogs cannot be photographed
- There is no problem with imbalance
Answer: A. The model may lean towards predicting “cat” and do worse on dogs
Hint: Which class did the model see much more often?
- The model sees cats 9 times more often.
- It can score well overall just by guessing cat a lot.
- Collecting more dog photos balances the data.
Skill: Decomposition · Level: medium · Report an error
Q3A model is 95% accurate for one group of users and 70% for another. What should the team conclude?
- The model is unfair to the second group and needs investigation
- An overall average is all that matters
- The second group is using it wrong
- Nothing; 70% is the same as 95%
Answer: A. The model is unfair to the second group and needs investigation
Hint: Compare the groups, not just the average.
- A high overall score can hide poor results for one group.
- A 25-point gap means the second group gets many more errors.
- The team should check the data for that group and improve it.
Skill: Decomposition · Level: challenge · Report an error
Q4Many diseased leaves in a dataset were labelled “healthy” by mistake. What happens to the model?
- It learns wrong patterns and may call sick plants healthy
- It fixes the labels itself
- Nothing changes
- It becomes more accurate
Answer: A. It learns wrong patterns and may call sick plants healthy
Hint: A model treats labels as the truth.
- The model learns from labels as if they are correct.
- Wrong labels teach it that disease signs mean “healthy”.
- Checking labels during data exploration prevents this.
Skill: Abstraction · Level: medium · Report an error
Q5What does “consent” mean when collecting data?
- People agree to their data being collected and used
- Data is collected secretly
- Data is copied from anywhere
- Data is sold without asking
Answer: A. People agree to their data being collected and used
Hint: Think about asking permission.
- Consent means a person knowingly says yes.
- In India, children’s data needs a parent’s consent under the DPDP Act.
- Fair data is collected with consent.
Skill: Abstraction · Level: easy · Report an error
Q6A survey about how families use the internet is done only through an online form. Who is missing?
- Families without internet access
- Families with fast internet
- Nobody
- Only teachers
Answer: A. Families without internet access
Hint: Who could not fill in an online form?
- An online form only reaches people who are online.
- Families without internet cannot answer.
- So the results do not represent all families.
Skill: Decomposition · Level: medium · Report an error
Frequently asked questions
What does the Class 8 Data and Fairness AI chapter cover?
The Class 8 Data and Fairness chapter shows that an AI is only as fair as its data. Students check datasets for balance, representation, labelling errors and consent, and compare a model’s accuracy across groups to judge whether it is fair.
Which computational thinking skills does Data and Fairness use?
Data and Fairness mainly practises decomposition, abstraction. It links to Part 2: Artificial Intelligence.
Where are the official CBSE Data and Fairness questions?
The official questions are in the CBSE Class 8 Computational Thinking and AI handbook for 2026-27, free on the CBSE Academic website. The practice on this page is original and mapped to the same chapter.