From Marginal Revolution:

A new meta-analysis (pdf) of 72 studies, involving 4,795 groups and over 17,000 individuals has shown that groups tend to spend most of their time discussing the information shared by members, which is therefore redundant, rather than discussing information known only to one or a minority of members. This is important because those groups that do share unique information tend to make better decisions.

Another important factor is how much group members talk to each other. Ironically, Jessica Mesmer-Magnus and Leslie DeChurch found that groups that talked more tended to share less unique information.

A result that shouldn't surprise this group. I've noticed obvious attempts to avoid this tendency in Less Wrong (for instance, Yvain's avoiding further Christian-bashing). We've had at least one post asking specifically for information that was unique. And I don't know about the rest of you, but I've already had plenty of new food for thought on Less Wrong.

But are we tapping the full potential? Each of us has, or should have, a secret identity. The nice thing about those identities is that they give us access to unique knowledge. We've been asked (though I can't find the link) to avoid large posts applying learned rationality techniques to controversial topics, for fear of killing minds, which seems reasonable to me. Is there a better way to allow discipline-specific knowledge to be shared among Less Wrong readers without setting off our politicosensors? It seems beneficial not only for improved rationality training, but also to enhance our secret identities. For instance, I, as an economist-in-training, would like to know not just what an anthropologist can tell me, but what a Bayesian-trained anthropologist can tell me.