Help the user understand what the AI system is capable of doing.
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.