Ensure the AI system’s language and behaviors do not reinforce undesirable and unfair stereotypes and biases.
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.