AI and Fairness in Decisions
Explore whether AI should make important decisions about people, including decisions about jobs, loans and justice. Learn who is responsible when AI makes unfair decisions and the basics of algorithmic fairness.
What a learner can do afterwards
- Give an example of an important decision that AI might help make (hiring, medical diagnosis, loan approval)
- Explain why letting AI make decisions about people can be risky if the system is biased
- Suggest who should be responsible if an AI system makes an unfair decision
1 · Read
You have learned how AI bias can affect groups. AI can help choose interview candidates, assess loan applications or spot warning signs in medical scans. These are high-stakes decisions: mistakes can seriously affect people's lives. Choosing a playlist song usually has much smaller effects.
Imagine 100 job applicants in Group A and another 100 in Group B. Both groups are equally qualified. A tool invites 70 from A and 20 from B. This gap is a warning sign to investigate. It does not prove what caused the difference.
A biased tool can reject suitable people by copying unfair patterns. In a made-up case, two areas have equal crime levels, but one has more arrests because it has more policing. Treating the arrest gap as proof of more crime can lead to unfair recommendations.
People who build and use AI have responsibilities to check and address unfair results. An organization cannot avoid that duty by saying 'the computer decided'. In a hospital, a qualified health professional should review an AI warning before it guides treatment. People need a way to question and correct unfair decisions.
To compare counts, subtract the smaller from the larger. Our interview gap is 70 minus 20, or 50 more invitations. A count difference measures a gap. Judging fairness also needs information about qualifications, how decisions were made and who was affected.
Investigate unequal results in important decisions. People using AI remain responsible for checking and addressing unfair treatment.
Lesson, questions and flashcards written for this dataset. Applicant counts and the policing scenario are hypothetical teaching examples.
The reading was independently written for this dataset. The three linked videos are Code.org productions indexed by Khan Academy. Their publisher license terms have not been verified.
More to read
A review needs someone who can act
Imagine a tool rejects an interview application because a document is missing. The applicant explains that the upload failed. The missing file does not establish that the applicant lacks the skill it describes.
A reviewer should check the relevant information and correct the mistaken record. The organization should also investigate whether the upload problem affected other applicants.
Correcting one decision helps that applicant. Checking the process can reveal a problem affecting many people.
A review route needs someone able to change an unfair result. A complaints form alone cannot do that if nobody investigates or acts on what people report.
Explore more
2 · Watch
En anglaisTake it off screen
Where it sits
Learn first
Optional connections
This opens up
Where this leads
Jobs that lean on this skill. Follow one to see everything it is built on.
8 questions wait behind this lesson, each with its answer explained. Every answer feeds the sky: stars light as they are learned, and dim when it is time to come back.