Engage in disambiguation or gracefully degrade the AI system’s services when uncertain about a user’s goals.
Enable the user to provide feedback indicating their preferences during regular interaction with the AI system.
Immediately update or convey how user actions will impact future behaviors of the AI system.