AI Ethics & Governance

Bias, fairness metrics, EU AI Act, watermarking, and responsible deployment.

Systematic errors in a model's outputs that unfairly advantage or disadvantage particular groups, often reflecting biases in training data.

e.g. A hiring model trained on historical data disadvantaging women applicants.

A mathematical guarantee that adding or removing any single record barely changes a model's output distribution, protecting individual privacy.

e.g. Adding calibrated noise during training so no one person's data can be reverse-engineered.

A facially neutral policy or practice that disproportionately harms a protected group, even without intent to discriminate.

e.g. A loan-scoring rule that unintentionally denies a protected group at a higher rate.

The European Union's binding law that classifies AI systems by risk level and sets requirements scaled to that risk.

e.g. Banning real-time biometric surveillance in public spaces for law enforcement, with narrow exceptions.

The ability to understand and communicate why a model produced a particular output or decision.

e.g. Highlighting which words in a resume most influenced a hiring model's score.

A formal criterion, such as demographic parity or equalized odds, used to check whether a model treats different groups equitably.

e.g. Checking that a loan model approves applicants from different groups at similar rates.