Help the user understand how often the AI system may make mistakes.
Display information relevant to the user’s current task and environment.
Make it easy to invoke or request the AI system’s services when needed.
Limit disruptive changes when updating and adapting the AI system’s behaviors.