@Fleur_Prinsen @jennymackness @Downes
Fleur, sure, we do need to incorporate Latour's 'non-human' agency too.
(So that we can move beyond the human/non-human dichotomy, and even more importantly, move completely away from the rather fashionable 'post-human' framework, which I think just tries to insert the 'human' back into the centre of the debate by feigning denial!)
One way is to go with avatars, broadly speaking, in which (in a conversation prompted by Sian Bayne's presentation at Sterling some time back) we explored the idea that a thesis (a PhD, in particular) can be seen as an avatar of the writer - a 'proxy of a special kind', which is let loose to make its own way (somewhat independent of the writer) in the big wide world.
Another way (prompted by a presentation of Stephen Downes's), is to go with probes, see:
Probes are "experimental tools / ideas / interactive affordances; uncertain, tentative, interventions, that are inserted-and-let-loose within an emergent event", to see what happens (from elsewhere in this wiki, but it's directly in line with Snowden's practice, and writing).
Probes, specifically, respect (and assume) the unpredictable nature of a learning event, and the autonomous agency of the participants, and celebrate creative, surprising, unforeseen outcomes
(see https://learning-affordances.wikispaces.com/Probes for more).
Probes straddle the domains of human and non-human agency - for me - and its why I think of social learning tools - like aggregators, and footprints of emergence (https://footprints-of-emergence.wikispaces.com/) as probes - and I find it useful (not everybody does) to 'back load' probes with lots of the formal stuff on complex adaptive systems (but that's just me).
So ... when I think of designing (for) social learning, I use probes as a key metaphor.
And .. in work that I did with @jennymackness and Jutta Pauchenwein, our back-translation (from the German) prompted quite a useful description of 'autonomy' as 'independent initiative'.