Coordination Economies
I got a fairly insightful email asking why I called tags coordination machines, instead of search machines. The following is a dense and geeky explanation why - best skipped if you're not heavily with tags, I think.
Let's start by basically examining coordination. We can coordinate two things - preferences, and actions. The former is what we might call weak coordination - not a formal economic term, my own coinage, to refer to the fact that value creation potential is lower than the latter. I think it's intuitive that coordinating people's actions will result, on average, in greater value creation.
Now let's get back to tags. Tags are implicit preference coordination machines. They let people aggregate and disaggregate preferences multidimensionally. This is why they're powerful: if they didn't coordinate preferences, but only aggregated peoples' categorizations, we'd see explosions of useless tags.
That's not what the evidence suggests. Instead, people tag things they prefer, and ignore things they think don't prefer. In this way, preferences get implicitly aggregated, and then coordinated. That is, if 400 people tag bubblegen with technology, you might also reasonably suspect a significant proportion of them think it was cool enough to tag, which goes on to coordinate
your preferences about bubblegen.
Another example of a simple preference coordination machine is a hyperlink, but they're much weaker than tags, because they can only coordinate 1 thing with another (ie, a 1 to 1 relationship). Even so, we've seen (blogdex, etc) how powerful link aggregation is in terms of preference coordination. Tag aggregation will be a lot, lot more powerful.