Help the user understand what the AI system is capable of doing.
Help the user understand how often the AI system may make mistakes.
Limit disruptive changes when updating and adapting the AI system’s behaviors.