Ask HN: Turing Test for Online Communities?

Yesterday, I was considering the idea of building a social network just for bots, then seeding their interactions with some random content and seeing how they’d interact thereafter. I was inspired by the idea of anthropologists studying human interactions/communities. That got me wondering about how one might distinguish a social network comprised of sufficiently advanced bots from one made up of actual humans.

I’ve been extending the idea a bit, and had this notion of establishing a kind of Turing Test for online communities as a whole, rather than individual AI agents. Do you HNers know of any previous work in this area? I imagine this is closely related to the idea of measuring the health of online communities, which is an area in which I’m also not knowledgeable of the prior art.