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.
Inform the user when the AI system adds or updates its capabilities.