Bret Devereaux

"[Peter Turchin's theories] have won him comparisons to other authors of “megahistories,” such as Jared Diamond and Yuval Noah Harari." - From The Atlantic 'Megahistories' - Definition: Pop historical works that sound great unless you actually know enough history to assess them.

As @Peter_Turchin just said a few minutes ago, he's "just scientist". If you read Ages of Discord (or Ultra Society) he provides tools for seeing dynamic systems in histories. Then you can draw your own megahistory narratives.

But the issue is with the underlying data itself. The temptation in so many of these 'megahistories' is to develop a narrative based on personal biases or on very recent history, which is then simply projected backwards over vast periods for which the data doesn't exist.

If the data that sits beneath the model is broken, then the model itself is equally broken. Take for comparison the 538 guys and their election model. They are aggressively open about how, if the polls are wrong, then the model will also be wrong...

...consequently, they spend a lot of time stressing uncertainty. Turchin doesn't stress uncertainty *at all* in his comments. He stresses confidence, arguing that the huge date-brackets on his study give him the kind of high confidence to make hard future predictions.

And he openly compares himself to the historical-mathematicians from the Foundation books. What I am saying is - I *know* the data from some of these periods and, if you'll pardon the analogy - the polls are garbage.

And with evidence that is quite frankly so difficult, so contradictory, and so frequently weak or non-existent, his entire method strikes me as deeply uncertain. So where is that uncertainty? Where is the epistemic humility in his approach?