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.
Engage in disambiguation or gracefully degrade the AI system’s services when uncertain about a user’s goals.