> If you argue against a bot about a political matter, not knowing it’s a bot - you always lose on a long enough timeline. i.e. you can never persuade the bot, but it can wear you down / eventually find an argument that works.

That's a useful insight. The one on one wearing-down process came from theology. It's sometimes called Jesuitry. Members of the Society of Jesus were trained to do that sort of thing, and they got so good at it that the whole organization was suppressed in the 1700s.[1] There are still Jesuit-run schools in the US, and they do tend in that direction. "Never argue with a Jesuit", said Richard Nixon.

Since mass media, large scale use of one on one convincing has gone out of fashion. It's not cost-effective. With large language models, it's back.

[1] https://en.wikipedia.org/wiki/Suppression_of_the_Society_of_...