Purpose: To prevent AI systems/tools replacing or eroding human expertise, judgement and responsibility, and ensuring that automation supports human capabilities rather than diminishing them over time.
| Organisational / Technical | Measure |
|---|---|
| A. Defining legitimate boundaries | |
| ORG | The organisation has clearly defined what types of content, behaviour or expression are restricted and why |
| ORG | Restrictions can be justified on a legal, ethical or safety basis rather than reputational or convenience of the organisation |
| BOTH | The scope of restrictions is documented and communicated to relevant users or stakeholders |
| B. Human review and contestability | |
| BOTH | Humans can review, override or correct automated restrictions on expression |
| ORG | Clear mechanisms for affected individuals to report, question or contest moderation or restriction |
| ORG | Human review involves consideration of contextual, cultural or situational factors that automated systems may have missed |
| C. Transparency and monitoring | |
| ORG | Users are informed when and why expression is restricted using clear and accessible explanations |
| BOTH | The organisation monitors patterns of restriction to detect systemic over-censorship or bias |
| ORG | Policies governing expression and moderation are periodically reviewed as contexts, risks or norms evolve |
Source: AIOLIA deliverable 3.1