Enable the user to access an explanation of why the AI system behaved as it did.
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.