AP CSP topic 5.3
Computing Bias
Computing systems can reflect the biases of the people who build them or of the data used to train them. A face recognition system trained mostly on one group may work poorly on others, and a hiring program trained on past decisions may repeat past unfairness. Bias can enter at any stage, from collecting data to designing the algorithm. Testing with diverse data and diverse teams helps reduce it.
Computing Bias vocabulary
- Bias: A tendency that unfairly favors or harms certain groups.
- Training data: Data used to teach a machine learning system how to make decisions.
- Machine learning: A way for computers to find patterns in data and improve without being explicitly programmed.
- Representative data: Data that fairly reflects all the groups a system will be used on.
AP-style practice questions
Try each one, then open the answer. To save your progress, take this quiz in the interactive course.
A company trains a program to screen job applications using ten years of its past hiring decisions. In the past, the company hired very few women for engineering roles.
1. Which of the following is most likely to happen?
- (A) The program will hire more women to balance the past
- (B) The program may learn and repeat the bias against women
- (C) The program will have no bias because computers are neutral
- (D) The program will reject every applicant
Show answer
(B) The program may learn and repeat the bias against women A system trained on biased decisions can repeat that bias.
2. A face recognition system works well for some groups but poorly for others. Which of the following is the most likely cause?
- (A) The training data did not include enough images of all groups
- (B) The cameras were too expensive
- (C) The program was written in the wrong language
- (D) The system ran on too many processors
Show answer
(A) The training data did not include enough images of all groups Unrepresentative training data leads to uneven accuracy.
3. Which of the following is the best way to reduce bias in a computing system?
- (A) Test it with diverse data and include people with diverse backgrounds on the team
- (B) Use less data
- (C) Hide how the system works
- (D) Release it quickly and fix problems later
Show answer
(A) Test it with diverse data and include people with diverse backgrounds on the team Diverse data and teams help find and reduce bias.
4. Which of the following statements about bias in computing is true?
- (A) Computers cannot be biased because they follow instructions
- (B) Bias can come from the data used or from choices made by the people who design the system
- (C) Bias only happens in old programs
- (D) Bias only affects the programmers
Show answer
(B) Bias can come from the data used or from choices made by the people who design the system Bias can enter through data or design decisions.