I thought the best argument to treating virtual assistant politely is because virtual assistant is designed to be anthropomorphic, i.e. human like. You interact with your virtual assistant as if you are talking to another person. Therefore it (or she) deserves as much respect and politeness as another person, EVEN IF that person does not understand, like if you are hosting a foreign visitor who has little knowledge of our culture.
There are a lot of clumsy non-anthropomorphic assistants around, Google is the other kind of assistant using keywords. The most anthropomorphic assistant, of course, is the human butler.
By the same token, if someone treats everybody with contempt, it is hardly likely or beneficial to make that person interact with virtual assistant politely. Maybe there should be an “asshole” mode instead of “polite” mode, and that mode can be toggled on and off, and let users choose. Give consumers choice. Don’t enforce politeness or use default bias to nudge people towards one mode. This is the libertarian answer.