Correcting human error vs doing exactly what you're told - is there literature on this in context of general system design? — LessWrong