Help the user understand how often the AI system may make mistakes.
Immediately update or convey how user actions will impact future behaviors of the AI system.
Allow the user to globally customize what the AI system monitors and how it behaves.