I’ve been doing a fair amount with ChatGPT recently, and I do say please and thank you. Not because I think it has feelings, but because I think it’s important to be the kind of person who says please and thank you.
It seems analogous to the objection Puritan preachers had against bear baiting and dog fights and such. It wasn’t so much that they were concerned for the animals. They were concerned for the people who were watching.
I think it would be a reasonable position to say that the “kind of person” who does that can also be the “kind of person” who deeply values being gracious toward other humans but finds it totally unnecessary with machines.
However, there’s another wrinkle. You might want the AI to be exposed to human courtesy since it’s trying to emulate human speech and thought, and using courtesies with it would teach it to do that.
The idea that the language used in your questions might be training the AI is an interesting perspective.