Folksonomy Schmolksonomy

The tagging battles rage on, with a slew of posts by Clay Shirky provoking an unusually active discussion among the Many-to-Many blog readership. As a card-carrying acolyte of knowledge management who is deeply skeptical about current efforts to create a semantic web, I feel like I can sympathize with those on both sides of the debate.

On the one hand, Clay has a point. After years of effort by slews of brainy individuals, the semantic web is going nowhere. Then one day someone strikes upon the right balance of usability and utility to create a tagging system that is both simple enough for people actually to use and robust enough to be of value to a community of users, not just the one doing the tagging. The new Technorati Tags feature drives home the point raised by the brilliant-in-their-simplicity tagging systems used by del.icio.us and Flikr, showing how cool it is to unify them with each other and with blog tags (which have been around for a while in a non-folksonomy manifestion).

So now we have something that actually adds value, and something is undeniably a whole hell of a lot better than nothing. The last thing we need is for some pedant to crawl out of the woodwork and demand that we pile complexity onto a working (albeit somewhat rough-and-ready) solution in the service of an abstract academic goal that patently isn’t feasible anyway. The only way to avert disaster, runs this argument, is to push back against the ontology-heads with all our might until they crawl back into their ivory towers and let us get on with our business.

Certainly the fact that these tagging systems have costs as well benefits is well worth remembering. But Clay goes a bit overboard in his excoriation of ontologies. The most glaring flaw in his argument is the assumption that ontologies for tagging stuff on the web are going to look more or less like the Dewey Decimal System. There’s a long continuum between flat tagging and a bloated, rigid hierarchy designed to catalog physical objects which, by their nature, can only occupy one spot in the taxonomy. It seems to me that we have every reason to believe that the sweet spot lies somewhere in the middle of this continuum, not at the edge.

Clay uses abstract concepts like “creativity” to demonstrate the absurdity of trying to determine, a priori, where a given tag should be placed in a hierarchy that is unavoidably incomplete and subject to change in the future. Fair enough. But why throw the baby out with the bathwater? Hierarchy would also let me say things like “dogs and cats are both animals.” Then I could see pictures of dogs, cats, rabbits and hedgehogs on Flikr by keying them indirectly to the animal tag. That sounds useful to me, and since it seems unlikely that dogs and cats are going to stop being animals any time soon, Clay’s argument loses a lot of its punch.

Tim Bray also has a good point when he notes that namespaces would give more structured communities the option of creating more powerful tagging systems without mucking up the communal tagosphere. (Blimey, did I just coin a term!?) This is another way to introduce hierarchy into the mix without spoiling the unsullied elegance of the plain vanilla folksonomy.

Like it or not, I’m convinced that folksonomies are going to need hierarchy down the road if they are going to spread beyond the walled gardens where they currently reside. This doesn’t have to be an all or nothing proposition. I would favor a system that would let power users place their tags into a hierarchy if they want to without requiring this. So if I feel so inclined, I can create an apple tag and place it under the fruit tag so that referencing it doesn’t result in a couple of snaps of fleshy pomes buried underneath an avalanche of way cool Macintosh photos. When I add the apple tag to my photo/weblink/blog entry, the user interface should use a Google Suggest style drop down list to ask me whether I mean fruit/apple, computer/apple or just plain apple (or if I want to add a new hierarchical entry of my own). If multiple namespaces contain the tag, they could be proposed (clearly labeled, of course) in the list as well

If done correctly this would make adding the value of hierarchy to a tagging system fairly painless. Of course, “if done correctly” is a pretty big caveat, and this would be easy to get wrong. But the same applies to flat tagging, and this didn’t stop people from experimenting and innovating until they got it right. Assuming that it’s really as useful as I suspect it to be, I wouldn’t be so hasty to reject hierarchy either.

Bookmark to:

8 Comments

1. I agree completely that to make del.icio.us really powerful it needs other dimensions such as hierarchy. But in fact, del.icio.us has a rough & ready hierarchical capacity “built in.” I just use the MediaWiki convention of adding a colon after a main word and then name a subcategory. So I’ve got categories (or namespaces, whatever) and subcategories. It’s a personal hierarchy within a flat system.

   People I work with know my little system and can subscribe to those tags — only drawback right now with del.icio.us is that it doesn’t have multiple inboxes, so I can have multiple personally-tailored aggregated feeds. But I can achieve something similar by grouping related tag feeds in categories in my RSS readers.

   There are all sorts of things you can do with del.icio.us to personalize how it works for you and, with a small group, how the group shares info and actionable items.

   My first brainstorming is on my blog — Eureka, persononmies! but I’m finding new ways of doing things with my small group about five times a day right now. Hopefully the innovation pace will slow down a bit by next week — it’s making my head hurt sometimes!

   Overall, my main insight — with a bit more flexibility and a bit more power (e.g. multiple inboxes) — del.icio.us can do a fantastic amount of stuff — including building hierarchies and controlled vocabularies. But it’s bottom up built — it’s not that the system creates the structure, it just gives you the tools. You adapt them to your use needs.

   What I think Shirky’s really right about is that this is a use process, not a classification system. Not being an IA pro, I find the fundamental conservatism and “control freak” mentality of some quite amusing. Shirky just seems to be reacting to those in the IA field who are allergic to the idea that users might muck up their systems.

   Related point — del.icio.us doesn’t have to change much — most of the core features they’re working on cleaning up will create the foundation. Others can introduce services to power users that give them the means to manipulate their del.icio.us tags and feeds. Still others will come up with del.icio.us interfaces for the people who want a no-brainer interface for their friends & family blogs, etc.

   So there will be competition among “elaborators” to produce tools that help people use their del.icio.us and extract more user value out of their del.icio.us.

   That means del.icio.us itself becomes a utility and its development of functions should take the KISS approach. It should focus on scalability, reliability, and the core functions that make it possible for others to build complementary tools.

   What’s worrying me right now are the economics and permanency of the thing if everybody starts using it as a utility. e.g., do I have a backup capacity if a main chunk of my personal CMS is sitting on del.icio.us and they can’t handle the volumes?

   Comment by nadezhda — 1/26/2005 @ 7:00 pm

2. I think it’s obvious that folksonomic systems will see changes both due to a desire for specificity as you describe as well as simple social conventions that make particular tags for particular kinds of items de reigeur if not outright necessary to link-in properly to other systems sha

   Comment by Chris L — 1/28/2005 @ 7:43 pm

3. ring data. At the same time, a “little bit of hierarchy” is a lot like being “a little bit pregnant” and I have to side with Clay that the minute the process starts becoming a top-down initiative for classification is the minute that system is doomed to– if not failure– certainly a limited application.

   Comment by Chris L — 1/28/2005 @ 7:44 pm

4. I don’t think Shirky was excoriating ontology so much as hierarchy, which he mislabels as ontology, as Rick Thomas commented. Shirky’s point about strict hierarchy’s weakness (needing to predict the future) was valid, and tagging’s simple method to get to an object via several paths illustrates this.

   And yes, you can implement hierarchy via tags by simply having some tags being strict subsets of others, and it should be simple to extend the functionality of the tools to support this more easily.

   Comment by fling93 — 1/28/2005 @ 10:13 pm

5. You can create a kind of hierarchy by bookmarking the URLs of your del.icio.us tags themselves as well. I wrote about this in this post del.icio.us links and hierarchies.

   Comment by mcharper — 3/7/2005 @ 1:44 pm

6. Sorry, that should be here.

   Comment by mcharper — 3/7/2005 @ 1:49 pm

7. Folksonomy

   Folksonomy for Pottery Folksonomy is when people can append freely chosen keywords to organize data ( wikipedia .) As noted the other day, people are uploading pottery images to Flickr and organizing them.

   Trackback by John Norris — 4/28/2005 @ 3:50 pm

8. I have just implemented a folksonomy with hierarchy (3 level max at the moment), it is still pretty much in its infancy and potentially buggy, however if anyone wants to try it, it is at http://www.bigblogzoo.com.

   You use the tree to navigate to one of the blogs or feeds and then you press ‘animalise me’, which brings you to the folksonomy.

   If anyone has any comments I would be very interested to hear them:

   http://www.syndicatescape.com/bigblogzoo/

   Comment by Kent — 6/16/2005 @ 4:29 am

Trackback URL RSS feed for comments on this post.

Sorry, the comment form is closed at this time.