Differential Testing
A testing methodology that compares the behaviour of an AI model across different demographic groups or input conditions to detect inconsistencies and potential biases.
In plain language
Testing AI by comparing how it performs across different groups. Does the medical AI diagnose equally well for men and women? For different age groups? This testing reveals hidden inequalities.
Why this matters
Differential testing is a governance tool for ensuring AI fairness. Your framework should require testing across demographic groups and input conditions to detect disparities before AI systems are deployed to production.
Relevance
GovernanceDifferential testing is a mandatory governance control for validating that AI systems do not discriminate against protected groups, supporting compliance and ethical deployment standards.
Related terms
Putting differential testing 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.