Clay Shirky responds to Nick Carr

I don't think that's a digression at all; the various different
phenomena that go under that label are not merely different from one
another, they often have antithetical principles. Open source involves
a lot of talking (a LOT of talking), whereas jellybean guessing
requires that the participants not be talking to one another, to
maximize exposure of local knowledge. In fact, for significantly
important jellybeans (like the closing price of IBM) evidence that
talking occurred among the participants is a crime.

As an aside, part of the dilemma is that two recent books explaining
some kinds of group effects -- Wisdom of Crowds and Crowdsourcing --
stretched that word to mean both collaboration and it's opposite.

Others gain their usefulness by essentially filtering out those individual talents. Some crowds might be called "hives," which implies some degree of individual unconsciousness about how one's work or behavior fits into the larger whole, while others aren't anything like "hives."

Although I like this distinction, I don't think it's the presence of
unconscious behavior but rather the absence of conscious behavior that
makes the difference. Even fully intentional work groups are guided by
unconscious principles; it's the lack of such intention that makes
reality mining, say, different from Wikipedia.

To Nick's taxonomy -- social production crowd, averaging crowd, data
mine crowd, networking crowd -- I'll add a fifth, which is the
transactional crowd. Match.com, eBay, Innocentive, LinkedIn and
similar services use a social substrate to coordinate what are mainly
or solely point-to-point transactions.

I'll also add another complicating vector, which is scale. To take one
example, the vast majority of working social production groups operate
at collaborative scale (~3-5 people) where the imperative is to
increase communications among members, while for a tiny fraction of
large projects (like the Apache webserver, or LOSTpedia), the goal is
to _reduce_ the required collaboration, so that thousands or millions
of people can contribute.

Both kinds of groups are indisputably "social production", but small
groups seek out what would kill big groups and v-v. The same issues of
scale pervade networking groups, while datamine and averaging crowds
don't treat small and large differently, and generally benefit from
scale, as it gives them more data points.