Label Bias
Systematic errors in data annotations caused by the subjective judgments, cultural backgrounds or incentive structures of human annotators.
In plain language
When the humans labelling data bring their own biases to the task. If annotators consistently rate certain accents as "less professional" or certain demographic groups more negatively, the AI trained on that data will learn and amplify those biases.
Why this matters
Label bias is a data governance risk that directly affects AI fairness and performance across demographic groups. Your governance framework must include quality assurance for data labelling, including diverse annotator teams, clear labelling guidelines, inter-rater reliability checks and periodic audits of labelling consistency across demographic groups.
Relevance
GovernanceLabel bias can embed discriminatory patterns into AI systems, requiring governance controls over data annotation processes to ensure fairness.
Putting label bias into practice in your organisation?
Ready to transform your AI strategy?
Partner with Australia's AI strategy and governance specialists. From adoption roadmaps to ISO 42001 audit readiness.