As AI systems become more consequential, ethics reviews are becoming standard practice. From conducting reviews and flagging potential bias to documenting model impact and auditing decision-making, this work requires specific vocabulary. This exercise covers the collocations used in responsible AI frameworks, model card documentation, and AI governance meetings.
0 / 5 completed
1 / 5
The AI team was required to ___ an ethics review before deploying the new recommendation model to production.
Conduct an ethics review is the formal collocation — ethics reviews are 'conducted' as structured processes with defined participants and outputs. 'Complete' implies finishing a checklist; 'run' is informal; 'perform' is also acceptable. 'Conduct' is the standard verb in AI governance and responsible AI framework documentation.
2 / 5
The ethics board asked the team to ___ any potential bias in the training dataset before sign-off.
Flag potential bias is the natural AI ethics collocation — risks and concerns are 'flagged' to draw attention to issues that need investigation. 'Identify' is more analytical; 'report' and 'document' are subsequent steps. 'Flag' implies raising the issue proactively for review, which is the expected behaviour in AI ethics governance.
3 / 5
The team was asked to ___ the impact of the model on protected characteristics in the model card.
Document the impact is the standard model card and AI ethics collocation — model cards 'document' impact, limitations, and intended use. 'Assess' is the prior analytical step; 'describe' focuses on narration; 'analyse' is the method. 'Document' is the prescribed action in the model card template and AI ethics review checklists.
4 / 5
The responsible AI team ___ a set of guardrails to prevent the system from generating harmful outputs.
Implemented guardrails is the precise AI safety collocation — guardrails are 'implemented' as technical or policy controls. 'Built' is also natural for engineering teams; 'added' and 'created' are less specific. 'Implemented guardrails' is the standard phrasing in AI safety and responsible AI engineering discussions.
5 / 5
The ethics reviewer recommended that the team ___ the model's decision-making process to ensure explainability.
Audit the decision-making process is the formal AI ethics collocation — model audits provide independent verification of how decisions are reached. 'Review' is broader; 'examine' and 'inspect' are less formal. 'Audit' carries the connotation of external accountability and compliance, which is central to AI ethics governance frameworks.