In Lorenz the real result is a significant improvement of the collective estimate under social feedback, which is the opposite of what the title and wording in the report imply. By taking group diversity as an estimate of the residual error, it grows relative to that estimate, because conformity makes that diversity fall far more rapidly than the error. So the real results are a) crowd estimates get better with feedback and b) crowd diversity is no reliable measure or estimator for the error.
It also shows once again that the arithmetic mean and standard deviation are nearly always inappropriate statistics to use in the social sciences and that with rare exceptions the median and quartiles are far superior. But in practice the former is the one reported near exclusively while the latter rarely is.
There is a very good objection to the Neanderthal cooked-grain hypothesis by Collins and Copeland. But to my mind it’s comprehensively and competently refuted by Henry et al. So it seems that, unexpected as it was, the cooking result stands firm.
Under the sub-headline “Testing the Model Predictions” all McNerney et al. have to say is, that testing ought to be done. Until it is their model remains a nice but meaningless numerical exercise.
Here’s the link to this week’s complete list.