Help the user understand how often the AI system may make mistakes.
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.