Stakeholders and society HL only
Algorithmic bias
Also called AI bias
Where you meet this term: IB Diploma
Systematic unfairness in the results of a computer system, such as an AI model, that favours or disadvantages particular groups of people, often because it learned from biased or incomplete data.
In context
A system trained on past decisions can repeat past prejudice, for example in recruitment, lending or which customers receive the best offers. Because the results look objective and the workings are hard to see, bias can go unnoticed. Businesses reduce it by testing results across groups, improving their data and keeping people accountable for decisions.
Example
A recruitment tool trained on a firm's past hires keeps ranking applicants from one kind of university highest, because those are the people the firm hired before.
Related terms
Suggested next
- Big data Linked through Artificial intelligence (AI) and Artificial neural network (ANN) and Customer loyalty programme
- Database Linked through Data privacy and Customer loyalty programme
- Critical infrastructure Linked through Artificial neural network (ANN)
- Cybersecurity Linked through Data privacy
- Data analytics Linked through Artificial intelligence (AI)
- Data mining Linked through Customer loyalty programme