Your choice touches other people.
Check who bears the errors, whose information is used and whether the person affected can object without penalty.
- Dignity
- People are more than inputs or outputs. Check that depictions of people are respectful and accurate.
- Fairness
- Check who bears the errors when work is handed over. Protect those least able to object first.
- Privacy
- Do not give a system personal information without need and permission.
- Consent
- Tell affected people when AI is involved. Let them decline where offering that choice is reasonable.
- Autonomy
- Do not pressure anyone to accept a result because a tool made it.
- Transparency
- State where AI was used, and how.
- Accountability
- A named human answers for each stage. 'The system decided' is never an answer.
- Human override
- Any result can be overruled. Any process can be stopped, without penalising the person who stops it.
Could the person affected find out what happened, ask a human and have the outcome changed?