Can’t be bothered looking into it right now, but writing down for something to come back to.
A hypothesis, that flocking behaviour is so ingrained that humans will express it in their writing.
I did do a few Wikipedia pages when the thought crossed my mind, and apparently humans do flock like sheep/birds/wildebeasts – in a crowd, if 5% change direction, the rest of the mass will. But I’m suggesting it might run deeper.
Take a bunch of papers/blog posts or whatever, do similarity metrics, but hold on to the author as a node. Track them over time, I suspect when mapped down to lower dimensions will exhibit flocking characteristics. Could be interesting, where in bird flocks a predator comes along, splits the flock. In doc space that’s probably a new, potentially useful idea. Or not. Difference could be interesting.
It seems a lot easier to generate flocking behaviour (google Boids) than to recognise it. But if someone bothered to make the data pliable from mailing list archives or whatever, someone else bothered to do a visualization (graph nodes, distance, that’s an easy one), I reckon if there’s any validity to this hypothesis, it’ll be obvious to the eye.