Help the user understand what the AI system is capable of doing.
Ensure the AI system’s language and behaviors do not reinforce undesirable and unfair stereotypes and biases.
Engage in disambiguation or gracefully degrade the AI system’s services when uncertain about a user’s goals.