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.
Limit disruptive changes when updating and adapting the AI system’s behaviors.