I like your analysis before the "--".

It depends on (p) the probability of an uncivil and inarticulate post being correct. If p is very low, the risk of misplacing status outweighs the possible benefit of truth. If p is very high, vice versa.

What is your p? At what value of p is the boundary between the two policies?

(Allow me to treat "p(correct)" as "a general measure of (probable) goodness of content" so as to not be distracted.)

Both the appreciation of a post and the inclination to award status can (, usually are and probably ought to) be dependent on other factors than correctness. This may either be because having pleasant conversation (civility and articulateness) is a terminal value or because civility and articulateness are instrumental in achieving a goal beyond the scope of the immediate calculation. Either way, p = 1 can be easily dominated by other concerns. p is insufficient for the representation you are aiming for. Consider adding a direct weighting.

It depends on (p) the probability of an uncivil and inarticulate post being correct.

Again on this part, I am not sure whether you mean: * The prior probability of any particular post that is uncivil and inarticulate being correct is 0.83. * This is a post. It has 0.83 chance of being correct. It is also uncivil and inarticulate.

Should we priviledge a over b or vice versa?

I personally tend to weigh them approximately equally. At least, I can't guess which of the two I privilege more. (Again, I am using 'correct' extremely liberally.)