In complexity theory (see Assessing Complexity here, and also on the K-M-etaphors wiki) self-organisation and emerging properties are enhanced by a balance between some degree of freedom and some structure.

One of the key structural issues in designing learning 'ecologies' or 'virtual adaptive networks' is the issue of synchonisation - to what extent is learning a social / community practice, and how can you design learning to achieve this? Here are some thoughts on the issue:

Consumers or Contributors
Mike Caulfield recently wrote an interesting piece about a class-less society. He means of course school ‘class’, and he advocates for the use of ‘cohort’ instead, because of its associations with “’generational cohort’ … a group of people that experience a certain set of events simultaneously together.
He references John Seely Brown: too, and Stanley Fish:
“Together, members construct and negotiate a shared meaning, bringing the group along collectively rather than individually. In the process, they became what the literary critic Stanley Fish calls a “community of interpretation” working toward a shared understanding of the matter under discussion,” and what Tony Hirst calls OCW content delivered serially:.
Caulfield says …
“What you would need, ala Hirst, is a serialization mechanism (and here, again, talking in terms of the original meaning of serialization, not it’s specialized meaning). You and your friends sign up to watch the mid-90s series Earth 2. and it delivers you an episode a week.
“In other words, you become a cohort, moving through this series in sync so that everyone shares a similar interpretative environment. …The “class” is dead, as is the “audience”. Long live the cohort”.
From the perspective of my own practice, this is an interesting idea, however …

1. “Cohort” sounds too formal and too constrained, but what's in a name? My personal choice is Community of Inquiry.

2. Much more fundamental though: In designing "digital ecologies"/ "virtual adaptive networks" for learning, the crucial shift is to move completely away from content-driven events to activity-driven events.

This gets us to the starting blocks for connectivist / inquiry-based / problem-based / activity-based learning /networked / CoP or workshop -based learning. Pick a term and an approach to suit your needs, but this is the threshold for learning now, surely?

So the serialisation mechanism is often required - it is a key driver for community, and learning is social and contextual, in many ways, no?

But ...
Focus serialisation on what the 'community' / cohort will DO, not what they CONSUME - sorry to put it so starkly, but I am not sure there is that much 'grey' left in this issue.