Ethics, Safety and Societal Impact
What this domain tests
Ethics, Safety and Societal Impact tests whether you can identify stakeholders, harms and safeguards across the AI and robotics lifecycle. Expect scenario questions on bias, privacy, consent, transparency, accountability, safety and human oversight.
Core rules and concepts
- Fairness means avoiding unjustified disparate outcomes; bias can enter through data, labels, features, objectives and deployment context.
- Privacy covers data minimization, consent, purpose limitation, security and retention. Anonymization is not automatically safe if re-identification is possible.
- Transparency and explainability let people understand and challenge decisions. Accountability assigns responsibility for design, deployment and monitoring.
- Safety includes fail-safe behaviour, testing, limits and human override, especially where robots share space with people.
- Human oversight matters for high-stakes decisions; automation bias is the tendency to trust a system too readily.
- Societal impacts include job change, accessibility, environmental cost and dual use. Compliance with a rule is not the same as being ethical.
Worked example
A school proposes facial-recognition attendance. Stakeholders include students, parents, teachers and administrators. Risks include consent and storage of biometric data, error rates differing across groups, and function creep into behaviour monitoring. Safeguards could include a non-biometric alternative, data minimization, published error testing, short retention, human review of disputed records and an appeals route. The reasoning is to reduce harm before deployment, not merely to add a disclaimer.
Traps that cost marks
Do not treat ethics as a final checklist item. Do not equate anonymity with privacy. Do not assume bias is only a data problem. Do not claim compliance alone proves ethical acceptability.