Help the user understand how often the AI system may make mistakes.
Limit disruptive changes when updating and adapting the AI system’s behaviors.
Allow the user to globally customize what the AI system monitors and how it behaves.