AI & Human Agency.

Prevent blind acceptance of AI recommendations

People accept a recommendation because it came from the system.

Encourage appropriate human review

High-stakes recommendations are followed with no review.

Detect inappropriate reliance

Teams rely on the system for calls they should still make.

Design effective human checkpoints

Checkpoints exist, but people click through them.

Detect excessive AI dependence

Work stops when the system is unavailable because people no longer do it themselves.

Preserve human expertise

Skill fades because the system does the hard part every time.

Maintain independent decision-making

People stop forming a view before they see the system answer.

Prevent deskilling

Roles lose the practice needed to notice when the system is wrong.

Define human/AI roles

It is unclear which steps belong to a person and which belong to the system.

Optimize task delegation

Work is handed to the system that a person still does better, or the reverse.

Design effective handoffs

Context is lost when work passes between a person and the system.

Coordinate human and AI decisions

A person and the system decide in parallel and contradict each other.

Preserve critical thinking

People stop checking reasoning once the system drafts the answer.

Understand effects on learning

Learners use the system in ways that skip the learning.

Prevent cognitive offloading harms

Offloading routine thinking also offloads judgement people still need.

Maintain human judgement

The final call is effectively the system call, and a person only signs it.

Preserve meaningful choice

The interface funnels people into the option the system prefers.

Prevent coercive AI interaction

Prompts, timing, or tone push a person past a choice they would have refused.

Enable informed decisions

People decide without the facts the system already has.

Prevent AI from restricting user agency

The system narrows what a person can see, say, or do.