Artificial intelligence gives the question of participation a practical setting. We can delegate more operations, compare more alternatives and work with larger bodies of material. The direction of that work still needs to be examined.
What can be delegated?
The Foundation distinguishes operations, the production of differences and preliminary evaluations. Searching for discrepancies, finding counterarguments, classifying material or comparing concepts can all be assisted by a system.
Delegation at these levels can expand a person’s possibilities. The boundary of agency does not run around every act of discernment. It concerns authorship of the direction and of the criteria by which that direction can be revised.
The person need not have produced every element of a decision alone. Direction takes shape in relations. Authorship involves recognising a direction as one’s own, answering for its continuation and retaining the capacity to reconsider its criteria.
Remembering facts, preserving the structure of an argument and deciding which parts of a life matter are different activities. Delegating one need not settle the others.
The last click
Keeping a final approval step is useful as an arrangement of responsibility. It does not settle the philosophical question. A person can approve a result without examining the goal, the assumptions or the standards that made it appear desirable.
Equally, a system may carry out many operations while the person remains actively involved in determining what matters. The number of delegated steps alone does not tell us how agency is distributed in an actual situation.
Support and dependence
A system can be valuable as an instrument for a long time. That value does not require treating it as the source of every criterion. The Foundation distinguishes a useful instrument from support that is intended to make a capacity available beyond itself.
The question is not how little AI a person uses. It is whether the terms of using it remain available to examination: which work is entrusted to it, what counts as a good result, what changes the direction, and what makes a decision revisable.
These distinctions do not establish the effects of any AI product. Research would need to examine actual tasks, alternatives, mistakes, changes of criteria and the person’s ability to revise delegation.