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.

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.

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.