Make it easy to dismiss or ignore undesired AI system services.
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.