Representation Bias
Bias arising when certain groups or perspectives are underrepresented or overrepresented in training data, leading to models that perform poorly for underrepresented populations.
In plain language
When some groups are underrepresented in training data. If a facial recognition AI is trained mostly on light-skinned faces, it will work poorly on darker-skinned faces. The AI inherits the imbalance in the data.
Why this matters
Representation bias is a governance priority because it directly causes AI systems to underperform for underrepresented groups and exposes your organisation to discrimination and fairness risks. Your data governance standards should require representation analysis and demographic performance testing before AI training data is approved for use.
Relevance
ImplementationBefore training or deploying any AI system, organisations must conduct demographic representation audits to identify which groups may be underserved, then establish performance thresholds and monitoring to ensure equity across populations.
Related terms
Putting representation 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.