Having written my post on cooperative distributed argumentation a while ago, I happened to stumble today on a (draft) paper by P.D. Magnus (thanks to Ektopos for the link):
The Promise and Perils of Science as Distributed Cognition
It's a fairly interesting discussion of (as you might expect) the distributed cognition model of science. One question I'd raise is about the specification of the purpose of a task: a purpose or function can be specified to various levels of precision. For instance, I can say that this keyboard's function is typing; this is a very vague specification. But someone could also give a more precise specification in terms of the particular construction of the keyboard, the sending of impulses of a certain sort to an input port, etc. So all a distributed cognition model of science would need to get off the ground is a very, very, very general specification of the function or purpose of scientific activity; the rest would be simply a matter of refinement. In any case, I thought I'd mention it, since I posted on a related issue recently.