← Back to course
AI Explorers 6-8 / Week 07 / Explorer Quest
6/6
Week 07 · Three Kinds of Bias

Explorer Quest

Explorer Quest: The Family Sample
// When a fair-looking score hides a gap
⏱ about 20 min

Explorer Quest: The Family Sample

Raven wants to know the family's favorite season, so she asks only her little brother.

"Winter!" he shouts, because he loves sledding. Raven writes "Winter wins" on the board.

Rocket laughs. "One person is not the whole family. That sample is tiny."

Nova floats over to the board. "So ask everyone," she says. "Then compare your two answers."

"Did your first sample tell the truth about the whole group, or only about one person?"

Raven erases the board and starts a fresh tally.

  1. Bias is a steady lean that makes results unfair or wrong for some cases.
  2. NIST names three kinds: systemic, computational and statistical, and human-cognitive.
  3. Each kind can happen without anyone meaning to be unfair.
  4. Non-representative samples are a common cause of computational and statistical bias.
  5. Check accuracy for each group, not just the overall score.
◇ FAMILY SAMPLE SURVEY
Pick a question with a few answers, such as favorite season or favorite board game.
First, ask only one person. Write their answer as Sample A.
Next, ask every family member or friend you can. Tally all the answers as Sample B.
Compare: did Sample A match the most common answer in Sample B?
Talk about which sample represents the group better, and why.
Answer choiceSample A tallySample B tally
For a grown-up
Ask: when have you seen a decision made from a sample that was too small or too similar?
Ask: why might an AI tool trained on lopsided data work well for some people and poorly for others?
Remind your explorer to ask a grown-up before using any AI tool outside this course.
QUEST CHECK
  • Read the question.
  • Tap your answer.
Sample A asked 1 person. Sample B asked 6. Which is more likely to represent the family?
Raven only asked her brother, but she did not mean to be unfair. Was her sample still biased?
Which kind of bias comes from a non-representative sample?
What word describes a sample that includes every kind of case in fair amounts?
WHY THIS EXERCISEA representative sample gives every kind of case a fair voice.
← Friday