Class 8
Chapter 3: Data and Fairness in AI
An AI is only as good as its data. Data should be accurate, complete and representative of all the groups who will use the system. Otherwise the AI may give bad or unfair results.
Learning outcomes
- Identify how data quality affects AI outcomes
- Reason about fairness and representativeness
Activities
- Audit a sample dataset for fairness
Worked examples
Read through these first, then try the practice below.
Example 1 — Spot the issue
An AI for selecting school cricket captains is trained only on past men's-team data. Why is this a problem if girls' teams use it too?
Solution: The data does not represent girls' play, so the AI's choices may be unfair to them.
Example 2 — Improve the data
What could improve the AI from Example 1?
Solution: Train it on a balanced dataset that includes both boys' and girls' teams, plus diverse playing styles.
Self-do practice
Question 1 of 3 · Score 0/0True or False: A small, one-sided dataset is likely to make an AI unfair.
Sign in as a student in this class to save your progress.