AI Bias Worksheet for Kids
AI tip (Fair data, fairer AI): Bias usually comes from the training data: if a group is missing or shown in only one way, the model copies that. One-sided data: one age group, only ripe mangoes, only adult voices, only neat handwriting. Balanced data: many ages, colours, voices, regions and styles. Reduce bias: add varied examples, check who is missing, test on many groups, fix wrong labels and listen to complaints.
NameDateTime takenScore ___ / 40
- 1.True or false: mangoes both raw and ripe is balanced data (fairer for everyone). (a) True (b) False
- 2.🛠️ Does this reduce bias or add to it: test on one group only? (a) adds to bias (b) reduces bias
- 3.🗣️ A voice app understands adults but not children. What is the likely reason? (a) few child voices in data (b) it was a rainy day (c) children speak too fast (d) the phone was new
- 4.True or false: stories with doctors of every gender is balanced data (fairer for everyone). (a) True (b) False
- 5.🔍 Which is the odd one out? (a) skip testing (b) add more of the same (c) fix wrong labels (d) test on one group only
- 6.True or false: stories where every doctor is a man is one-sided data (likely to make a biased model). (a) False (b) True
- 7.🖼️ If an image generator always draws a scientist as a man, that shows… (a) a fair model (b) a broken screen (c) a strong battery (d) bias from data
- 8.True or false: “ignore complaints” adds to bias. (a) False (b) True
- 9.True or false: photos of people of many ages and skin tones is one-sided data (likely to make a biased model). (a) False (b) True
- 10.True or false: “fix wrong labels” adds to bias. (a) True (b) False
- 11.True or false: neat and messy writing from many people is one-sided data (likely to make a biased model). (a) True (b) False
- 12.True or false: stories where every doctor is a man is balanced data (fairer for everyone). (a) False (b) True
- 13.Which of these adds to bias? (a) ask different people (b) check who is missing (c) balance the data (d) add more of the same
- 14.True or false: a fruit sorter shown only ripe mangoes is balanced data (fairer for everyone). (a) True (b) False
- 15.🔍 Which is the odd one out? (a) test on many groups (b) ignore complaints (c) ask different people (d) add varied examples
- 16.Which of these helps reduce bias? (a) test on one group only (b) skip testing (c) copy old unfair choices (d) fix wrong labels
- 17.🔍 Which is the odd one out? (a) add more of the same (b) copy old unfair choices (c) use one group only (d) ask different people
- 18.Which of these adds to bias? (a) test on many groups (b) add varied examples (c) test on one group only (d) fix wrong labels
- 19.🛠️ What helps make an AI system fairer? (a) more varied data (b) one group only (c) ignoring mistakes (d) fewer tests
- 20.⚖️ One-sided or balanced data: a voice app trained only on adult voices? (a) one-sided data (b) balanced data
- 21.🛠️ Does this reduce bias or add to it: copy old unfair choices? (a) reduces bias (b) adds to bias
- 22.⚖️ One-sided or balanced data: a leaf app trained only on summer leaves? (a) one-sided data (b) balanced data
- 23.⚖️ Where does bias in an AI system usually come from? (a) its colour (b) the weather (c) its screen size (d) its training data
- 24.🔍 Which is the odd one out? (a) test on many groups (b) add more of the same (c) ignore complaints (d) skip testing
- 25.🔍 Which is the odd one out? (a) fix wrong labels (b) test on many groups (c) ask different people (d) test on one group only
- 26.True or false: stories with doctors of every gender is one-sided data (likely to make a biased model). (a) False (b) True
- 27.Which of these helps reduce bias? (a) skip testing (b) copy old unfair choices (c) ignore complaints (d) add varied examples
- 28.True or false: a voice app trained only on adult voices is balanced data (fairer for everyone). (a) False (b) True
- 29.True or false: “skip testing” adds to bias. (a) False (b) True
- 30.True or false: neat and messy writing from many people is balanced data (fairer for everyone). (a) True (b) False
- 31.⚖️ One-sided or balanced data: photos of people from only one age group? (a) balanced data (b) one-sided data
- 32.Which of these adds to bias? (a) copy old unfair choices (b) add varied examples (c) test on many groups (d) balance the data
- 33.True or false: leaves from every season of the year is one-sided data (likely to make a biased model). (a) True (b) False
- 34.⚖️ One-sided or balanced data: photos of people of many ages and skin tones? (a) one-sided data (b) balanced data
- 35.True or false: a fruit sorter shown only ripe mangoes is one-sided data (likely to make a biased model). (a) True (b) False
- 36.Which of these adds to bias? (a) add varied examples (b) ignore complaints (c) test on many groups (d) check who is missing
- 37.True or false: voices of children and adults from many regions is balanced data (fairer for everyone). (a) False (b) True
- 38.True or false: “test on many groups” adds to bias. (a) False (b) True
- 39.True or false: “add more of the same” helps reduce bias. (a) False (b) True
- 40.🛠️ Does this reduce bias or add to it: use one group only? (a) reduces bias (b) adds to bias
Answer key
- True
- adds to bias
- few child voices in data
- True
- fix wrong labels
- True
- bias from data
- True
- False
- False
- False
- False
- add more of the same
- False
- ignore complaints
- fix wrong labels
- ask different people
- test on one group only
- more varied data
- one-sided data
- adds to bias
- one-sided data
- its training data
- test on many groups
- test on one group only
- False
- add varied examples
- False
- True
- True
- one-sided data
- copy old unfair choices
- False
- balanced data
- True
- ignore complaints
- True
- False
- False
- adds to bias
Free ai for students lessons, practice and worksheets at talentjr.in/ai-for-students