This idea comes from anecdotal reports by friends on social media: it seems that once personal tastes and interests are narrowed down enough, even in larger urban areas, the people recommended by social media algorithms will converge.
That is not to say that you will be shown friends from other networks, but people that also know your friends from other networks.
This phenomenon has been discussed when measuring the human social network diameter (e.g., six degrees of separation) or the idea that the internat has transformed the world into a global village).
Here lies the question to be answered: do recommendation systems transform the global village into a small village when projected over a physical location? (The role of recommendation systems in humanizing an overpopulated world is a topic I have blogged before.)
This question can be addressed with some overall number of people, services, chances of people being on a particular service, etc. Or with a study using participants to measure their degree of separation **with respect to the people recommendations appearing on different services**.