[rnd0]

>Assume good faith initially because it least has a chance of being productive, but the moment bad faith is detected, either withdraw or "punish" (and propagate social signals that bad faith arguing has been detected so others know not to engage)

That's fair but I think a better solution is to enter into a conversation Tabla Rasa; you don't know someone's a kook/troll/whatever but you are aware that they may be. They may also be a potential friend/workmate/source of support/whatever.

With each interaction, over time, trust is either built up or torn down.

This is harder and harder to do in the current forum/anonymous posting climate not so much because of anonymity but because of different design elements that have stripped people's messages of ways to make distinctions (avatars, signature are two that come to mind, profiles are another).

Now posters all blend into one another so building up any kind of individual rapport is difficult; if not impossible.

[kelseyfrog]

Thanks for clarifying this. It made me reflect on the relationship between assuming positive intent, negative intent and blank slate as directly corresponding to strong positive and negative priors vs weak priors or even uninformative priors.

Even updating these priors with new data to form posteriors falls perfectly into place. Thank you.

[rnd0]

It's been good to have the chance to articulate my annoyance and examine it closer and think of what alternatives I'd propose. I'm glad to have something more to go into future discussions with than inarticulate gripes. All in all a net win IMO!