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.
Allow the user to globally customize what the AI system monitors and how it behaves.