I don’t think we need to make chatbots and voice interfaces more like people. We need to make them convenient and helpful.
Historically, we have skeuomorphism as a crutch to help people adopt new technologies and then cast it away if it isn’t pleasing AND functional. It first started in architecture (the Wikipedia article on it has some interesting past skeuomorphs).
I responded to another good article on voice interfaces here that you should probably check out: