There’s a useful and engaging discussion unfolding about “folksonomies” — emergent, user-shaped taxonomies of metadata like those in Flickr and Delicious (Adam Mathes’ thorough and detailed paper is here, Lou Rosenfeld offers measured dissent here, Clay Shirky fires back here). This topic reminds me of discussions we had back in 1999 and 2000, when we were building the Salon Directory.
We needed a tagging scheme for Salon articles, and some of the software developers felt that we should just generate a list of categories and build drop-down menus into the content management system. We had hired a smart consultant who argued that we should instead just let our editors add tags to stories in a free-form way, and allow the resulting categories to shape the “back catalog” of stories. As long as we occasionally did some gardening of the resulting keyword list — combining duplicate categories and handling complex issues (which “president bush”, exactly?) as they arose — we’d have a flexible, expandable schema naturally emerging from our daily work flow.
Our consultant was plainly right, and whatever problems the Salon Directory has had over the years have been more the result of limited software development resources on our part than of any fundamental mistakes in its conception. So within the confines of the Salon staff we had our own little “folksonomy” growing.
The biggest problems have not been those of organization, classification or structure but the simpler ones of time and effort. We made it relatively easy on our staff to add keywords, and some are added automatically, but it’s still a constant struggle to make sure every story is well keyworded. Some editors are more conscientious than others; all are on deadline much of the time; and the pressures of an “all-the-time” publishing schedule mean that today’s good intention of going back and fixing yesterday’s metadata failure usually falls prey to the demands of tomorrow’s stories.
I think this is what Shirky is getting at when he talks about how expensive it is to “build, maintain and enforce a controlled vocabulary.” Here’s in-the-field evidence that it’s not so cheap or easy to “build, maintain and enforce” even a within-the-firewall folksonomy. And so, much as I love the approaches of Flickr and Delicious, I also worry that the value of the tagging ecosystems emerging on those services will grow for a while and then, sadly, decline. Early adopters are enthusiastic and willing to take the time to tag; as the services grow, people are less likely to devote that time and care.
Which of course does not mean that these aren’t great projects — I agree with Shirky that they are far better than the alternative because the alternative, most often, is nothing at all. But when Rosenfeld and others wonder about the scalability of folksonomies, I think the issue may be less the scale of individual tags (50 billion “cat” photos!) than the scale of human enthusiasm for doing the slog-work of classification. Geeks love tidying up their personal dataspaces because, obviously, they’re geeks. For the rest of the world, my hunch is that — even when they’re only classifying the tiny sliver of stuff that’s their own — most people would rather do almost anything else.
Post Revisions:
There are no revisions for this post.