danah’s great piece on cultural issues in folksonomy gets to a key piece of the debate, namely that we can’t talk about categorization issues like accuracy without also talking about the culture that created the categories. However, I feel a curious disconnect between her exposition of the issues and her tone. There seems to be skepticism about folksonomic tagging in her post (though possibly it is just the reflexive skepticism of the academy.)

In any case, I want to point out that, for almost all the issues she raises, those characteristics are worse, much worse, in formal classification schemes than in folksonomies in general, and folksonomic tagging in particular.

At the risk of running good writing through the sausage grinder (and re-ordering it to boot) her list of issues is broadly:

1. There are perspective problems (e.g. the tag ‘me’ on Flickr.)
2. Tags can be gamed (in the manner of the MLK/Technorati tags)
3. Classification schemes are always culturally dependent
4. Many terms are contested
5. Some words cannot be simply translated literally
6. Some words have multiple or conflicting meanings

All the items on that list are true, and items 1 and 2 on the list are genuine design issues. System gaming is an issue, and can be fought with, inter alia, opacity of ranking method (the Google way), reputation markets (the Ebay way), continual post-hoc edits (the Wiki way), and so on. Each of these solutions may be tried in different places where folksonomy takes hold.

And for the relative tag problem, there may be a small enough number of those kind of tags — me, toread, unfiled, etc — that we can make a dictionary filter. But the relativity can also be interesting when crossed-tabbed with the identity of the tagger; I don’t want ‘toread’ or ‘funny’ generally, but I do want Liz’s ‘toread’ tags, and Matt Webb’s ‘funny’ links.

Items 3-6 on that list are different because while they are problematic in folksonomies, they are more — much more — problematic in top-down classification systems. Folksonomies represent progress in those areas, in other words.

You want cultural dependence? The Library of Congress, in its top level categories for geographic regions, lists “The Balkan Penninsula” as one main entity, and “Asia” as another. Contested terms? Try finding queer literature in any library classification scheme. And so on. Folksonomic tagging improves on this by exposing cultural dependence and contestedness, rather than denying its existence, or hiding it by fiat.

(As an aside, the signal loss from the pressures brought to bear on official categorizations is a common theme in classification generally. The entire alt. hierarchy in usenet came into being because there was a proposal to create rec.drugs, and there was concern that usenet, running in part over an NSF-funded network, would be shut down. The alt.* hierarchy was a compromise, to allow some face saving in suggesting that the *.drugs group was not ‘official’. And of course, alt. (an early folksonomy, albeit highly compromised by usenet’s hierarchical design) ballooned to many times the size of the ‘official’ usenet.)

The aggregate good of tags is not that they create consensus or accuracy; they observably don’t, and this is very observability is much of their value. Pick any popular del.icio.us link, click on the “and X other people” link under the URL, and you’ll see how that page is tagged by dozens or hundreds of people. There is both broad alignment around a few terms, but there is also a long tail of other views, which you don’t get in formal systems.

To take but one example, of the 114 people who tagged the Buffyology database, 66 tagged it buffy, 58 tagged it tv, and only 12 tagged it database, the third most popular tag. But 4 people tagged it sf or scifi, 2 tagged it fantasy, and 2 tagged it vampire. So the ambiguity between the literature of fantasy and of science fiction is exposed in the tagging, and the possibility of viewing Buffy as a thing related not mainly to TV but to vampires is also preserved.

So this is what I don’t get: I can’t imagine that anyone concerned about hegemony and marginalization would prefer professionally structured categories over folksonomy. If you care about contested terms and the risks of marginalization, del.icio.us, Flickr, et al do more to improve our access to, and understanding of, marginalization and contestation than any current alternative.