Glossary
Human Agency
Human agency is a person's capacity to make informed decisions of their own when using or being affected by an AI system.
Term: Human Agency
Human agency is a person’s capacity to make informed decisions of their own when using or being affected by an AI system. It depends on knowing AI is involved, understanding enough to judge its output and being able to choose differently or challenge the result.
It differs from AI agency, a system’s capacity to act autonomously.
Why it matters: agency and oversight
The EU High-Level Expert Group’s non-binding Ethics Guidelines for Trustworthy AI make ‘human agency and oversight’ their first requirement, with human oversight there to protect users’ informed, autonomous decisions. We place agency with the person using or affected by the system; oversight is the operator’s ability to monitor and intervene.
A system can be well overseen and still leave people with no real choice. Meaningful human review can take a decision outside UK GDPR Articles 22A to 22D on solely automated decision-making while the person affected may still receive no notice, explanation or route to contest.
How it works in practice
In its recruitment work, the ICO says candidates should understand when automation is used, the tool’s accuracy and the safeguards available; its updated ADM guidance remains in draft. For solely automated significant decisions, Article 22C requires information, the chance to make representations, human intervention and a way to contest.
Related terms: human in command, AI transparency and contestability and redress.
Want this in context?
See how this term fits into the broader programme of work.