Commenting on a recent post, Beth Harris asks the question of how the tagging system in Flickr could be used for teaching purposes. (Beth, a fellow SUNY-ite working at FIT, is doing some cool stuff with her art history classes using Flickr.) After thinking about it for a bit, I'm afraid the answer I come up with is "Not much; at least, not in its current form."
Given the huge buzz about folksonomies now, (which I have previously called distributed categorization), this is a disappointing discovery.
Here's my own take it:The problem with any sort of categorization scheme, whether it is generated through a top-down or a bottom-up process, is that there's no such thing as a universally useful scheme. We have a proliferation of metadata schemas in the top-down world (e.g., Dublin Core) because different schemas serve different purposes. It is therefore silly to expect that some magically emergent category system will spring forth fully formed, like Athena from Zeus' forehead, just because you're enabling a bottom-up process.
Broadly participatory systems like Flickr and del.icio.us (and now Technorati) are going to be used for such diverse purposes that you'll get a least-common-denominator system that is useful to nobody. This is not to say that you won't occasionally get a tagging convergence that is interesting or even useful; but these will tend to be exceptions that prove the rule. For the most part, folksonomies that are built up with such a broad audience for such diverse purposes are likely to be fun toys and not much more.
To get around this problem, some people are experimenting with creating specialized tag sets with a prefix. For example, art history professors and students using Flickr for class purposes could prefix all of their tags with "arthistory:", e.g., "arthistory:flemish". Using this system, Beth and her students could draw on only those tags that are being used for similar purposes to their own. In Geekspeak, they are hacking an ad-hoc namespace. Within this smaller world of tags, the students might have some hope of discovering, say, images of Flemish art that is contemporary to Merode's.
Two additional thoughts:
First, this strategy runs counter to the pervasive meme that all problems can be solved by harnessing network effects and growing some emergent...thingamabob. The thing about emergence is that it depends on the individual agents to be following a set of simple rules that effectively send ripples through the mob. But with wide-open folksonomies, individual participants may be using very different rule sets, depending on their individual aims. There's no particular reason to believe that folksonomies, in and of themselves, create the necessary preconditions for emergent order.
Second, I wonder if we're missing a useful property of folksonomies because of the way that they're being implemented. The emphasis right now is on discoverability of resources, i.e., using these tags to find stuff related to the stuff you're interested in but that you didn't previously know was out there. That makes sense, since (a) discoverability of resources is one of the great challenges of the Internet, and (b) since these tags are being used as categories for bookmarks, it is only natural to think of them in terms of organizing resources to make them more discoverable (or re-discoverable, if they are your own bookmarks).
But we may ultimately be able to mine more interesting stuff by looking in the opposite direction. Rather than checking to see what other resources share the same tag, how about looking at what other tags are attached to the resource? Tags, being metadata, tell us something about the context in which the tagger is placing the resource. So, going back to our art history example (but this time for research rather than teaching), it might be interesting to see how different historians are tagging the same images. We could learn something about how our colleagues contextualize individual artifacts for themselves.
Most interestingly, in this particular context it is precisely the mismatch of tags that yields the most knowledge. In any field of inquiry, the edge cases are where some of the most interesting work gets done. Is the painting late Impressionist or early Expressionist? (This may be a lousy example, since I am not an art historian.) This is a somewhat similar philosophy about the role of metadata to the one being taking by the APOLLO Learning Object Repository project, i.e., the object itself is at the center with various metadata bits tacked on extensibly. By putting the "meta" before the "data," maybe we're putting the cart before the horse.