Confirmation bias can seep into the process of data annotation, where humans label the data that AI systems learn from. If the annotators have preconceived notions about what the data should show, they may label data in a way that confirms their beliefs, inadvertently teaching the AI to reflect these biases.
- Log in or register to contribute
Contribute to three or more articles across any domain to qualify for the Contributor badge. Please check back tomorrow for updates on your progress.