Help the user understand how often the AI system may make mistakes.
Limit disruptive changes when updating and adapting the AI system’s behaviors.
Enable the user to provide feedback indicating their preferences during regular interaction with the AI system.