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