humans are not perfect Bayesian reasoners...
This seems like a selective application of a universal objection. "We almost always either adjust too far and reverse stupidity or don't discount enough," makes your reason technically not a universal objection, just an objection to learning anything from badly biased people, but this might be a confabulation added to the universal "humans are not perfect Bayesian reasoners".
There is a difference between suboptimal updating and updating in ways that don't correlate with the truth. My point is that badly biased posters may cause us to do the latter. But this is really tangential to the main thrust of my argument. Even (if) you can learn things (from) the occasional source Sam cites that seems very unlikely to make up for the rest of his noise. Well-Kept Gardens Die by Pacifism.
I wanted to bring attention to two posts from Razib Khan's Discover magazine gene expression blog (some of you may have been readers of the still active original gnxp) on the polemic surrounding Pinker's The Better Angels of Our Nature.
Relative Angels and absolute Demons (and the related But peace does reign! )
I generally agree with some of his arguments, but found this quote especially as summing up some of my own sentiments: