Enable the user to access an explanation of why the AI system behaved as it did.
Maintain short-term memory and allow the user to make efficient references to that memory.
Allow the user to globally customize what the AI system monitors and how it behaves.