Help the user understand how often the AI system may make mistakes.
Display information relevant to the user’s current task and environment.
Engage in disambiguation or gracefully degrade the AI system’s services when uncertain about a user’s goals.