Practice vocabulary for responsible AI development: responsible AI principles, fairness-aware ML, demographic parity, model performance on minority classes, and testing for bias.
0 / 5 completed
1 / 5
Your organization publishes a set of ___ AI principles covering fairness, transparency, and accountability.
Responsible AI principles are an organization's stated commitments to building and deploying AI ethically — typically covering fairness, reliability, safety, privacy, inclusivity, transparency, and accountability.
2 / 5
The ML team adopts ___-aware ML to ensure the model does not disadvantage any group.
Fairness-aware ML integrates fairness constraints and metrics into the model development process — choosing algorithms, loss functions, and evaluation criteria that reduce discriminatory outcomes.
3 / 5
An audit reveals the model violates ___ parity: approval rates differ significantly by protected group.
Demographic parity (also called statistical parity) requires that the proportion of positive predictions is equal across demographic groups. Violating it means the model favours or disadvantages certain groups.
4 / 5
The evaluation report notes: 'The model performs worse on ___ classes.' What does this mean?
When a model performs worse on minority classes, it means that underrepresented groups in the training data receive less accurate predictions — a common fairness problem caused by imbalanced datasets.
5 / 5
Before launch the team runs a suite of tests ___ for bias to verify fairness across subgroups.
Testing for bias involves running targeted evaluations that compare model performance across protected attributes (gender, ethnicity, age, etc.) to identify and quantify unfair outcomes before deployment.