With the exception of the Oxford example, the tradition is to pair students with one another and have a teacher listen in periodically. This does scale fairly well.
Hum, sound something similar to Montessori schooling but honestly while prized here and there so far I fail to see any real tangible effectiveness...
Sure, peer learning scale fairly well AND if one peer know can probably successfully transmit with a bit of supervision to spot and correct eventually transmitted mistakes and sure try teaching something is a nice way to self-test their own knowledge and so decide where to go from the current state to know more, but such models IMO are very slow and works only in very well crafted environments where nearly all peers have already a certain cultural background and a bit of habit. Something I do not see much present in the real current world.
Books just need printing capacity, logistic, reader interest witch are much more present on scale: printing books does not means a dedicated industry, printers and binders can works for various kind of papers product for much of the industrial processing. Logistic it's the same.
However analyzing the "network effect" of such models might be very interesting.
