The experiment I refered to (
https://forums.civfanatics.com/thre...ther-ai-survivor-spinoff.691131/post-16811568) does suggest that enabling barbs adds more randomness, since in the one case I had forgotten to disable them for roughly half the runs, the map's unbalance became more stark once they got disabled.
That's also something I seem to be observing in my new experiment (Horatius League): I mentioned a map where a start had a 28/30 win rate, which you never get to see with barbs on.
Another suggestion to that effect would come from the other end of the spectrum, when I tried to asses the
impact of the Raging Barbarians settings.
But so far, it's not "evidence", just a pointer.
What I have
not observed is a difference between the AIs in the way they handle barbarians. I also expected that turning them off would give an edge to Imperialistic leaders (barb cities fill the spot they could expand into), and that failed to materialize.
Barbs create randomness through purely random events (AI gets an unlucky roll and loses an early city or settler, barb city spawns early in a bad spot for the AI), but mainly through the randomness of the barb city captures (creates border tensions where there should be none, war with unreachable city scenario, etc.).
I intend to run my League 2.0 with barbs off as well, then we'll have a dataset we can truly compare to Keler's.