Joho the Blog
An Entry from the Archives

I've come across this argument a few times within IT depts who want to design well-managed information stores - "you need structure to help the users" - and I think it misses the point. An efficient index is unlikely to be useful if it doesn't contain the information you are looking for, and that is more likely to happen when you have to apply pre-determined tags that may be taken out of context. Rational designs rarely succeed in an irrational world. And I have yet to see a traditional classification system look anything other than perfectly rational...

I touched on this subject in a recent blog post: http://www.joiningdots.net/blog/2006/11/modernising-intranets.html

we have no practical choice but to use traditional taxonomies when organizing physical objects: The physical book can go on only one shelf

This is a non sequitur twice over. On one hand, you've already debunked the notion that imposing a single structure equates to imposing uniformity (Even Aristotle may have stored his clean togas in a different order than Plato without having to get into a metaphysical dispute). On the other, 'traditional taxonomies' don't necessarily choke on multiplicity - the systems used by librarians (boo, hiss!) can actually embrace duplication and multiple hierarchies without any difficulty (Peter Merholz made this point some time ago).

Taxonomies and folksonomies, and the activities that produce them, are different beasts - but we need to stop picturing the difference in terms of a great-chain-of-being hierarchy or a library shelf. Perhaps, as I suggested here, the only real difference between classification and ethnoclassification is that the former takes place within structured and credentialled communities.

Phil, although the physical book has to go in only one spot, you're obviously right that we can assign that spot multiple meanings. But, as I tried to point out with the card catalog example, when the assignment of those meanings also happens via physical objects--card catalogs, ledgers, etc.--there are practical limits on how many overlays of meaning we can manage.

Also, each overlay of meaning occurs via abstracted and greatly reduced metadata--what fits on a card or on a line of a ledger--and thus reduces meaning to make it more manageable. A single tag,of course, taken individually is as reductive as meaning gets. But, the tagging system overall can be quite rich with meaning. Further, the overall web of meaning within which online resources can exist can amplify meaning far beyond what a traditional taxonomy can manage.

(without reading the article--which sounds like a volley from the other side of the imaginary -onomies-arms-race:)

"good for the average user...since folksonomies will not produce an efficient index."

I am guessing this refers to the way an indexer / taxonomist / information architect / editor can use a hierarchy of concepts to show people less (= efficient) options at any one time.

I think one could argue that the (Yahoo! patented) clustering method in Flickr is also efficient in this sense of showing people less options at any one time. But, this brings up the larger issue with using the word "efficient" in this argument: efficient for doing what, exactly?

Strawmen, or in this case, straw-average-users, make for pretty weak arguments.

***

I also think it's sketchy to talk about taxonomies that "work" vs folksonomies that "work". For example, what Flickr works for is not necessarily applicable to other kinds of use.

This can be turned into an argument against taxonomies as well: what people need / use the structure for completely effects whether that structure is "right" or "works", regardless of the subject. If the structure is used for classification in the Aristoltean sense of mono-hierarhical classes, then it tends to work better if its a mono-hierarchical taxonomy :-)

Flickr also has an explicit taxonomy / ontology, and one could argue that this is what makes Flickr "work". But, again, it's the "works for what" / "efficient for what" question.

Elaine makes the argument that if an item on the web is tagged with words that do not describe it, then the system breaks down. In "The Anatomy of a Large-Scale Hypertextual Web Search Engine" by Sergey Brin and Lawrence Page the authors state, "Also, it is interesting to note that metadata efforts have largely failed with web search engines, because any text on the page which is not directly represented to the user is abused to manipulate search engines. There are even numerous companies which specialize in manipulating search engines for profit."

Metadata is data about data, and tagging a page on the internet is essentially adding metadata. For the same reason that search engines no longer rely on metadata, social bookmarking could be abused and eventually become worthless.

At a meta level (or is it meta-meta by this time) Elaine Peterson's argument is entirely reminiscent of the Wikipedia vs. Brittanica (false) dichotomy. In my view, I see such arguments as demonstrative of the old medium desperately attempting to hold onto its dominance long after it has been obsolesced (in the McLuhan sense of "no longer a structuring force in society"). The appeal to the putative pre-eminence of Aristotle literally (pun intended) characterizes ancient (that is primary literacy in the manuscript sense) thinking - not surprising for a librarianship prof.

Hmm ... here's the core of the debate, the claim mentioned above "If Weinberger means that it might be good for allowing individual users to supply their own tags, he might be correct. However, if he means that it will be good for the average user, his claim is questionable, since folksonomies will not produce an efficient index."

This is an algorithm-design problem in disguise.

Which is better:

1) An algorithm with performs very well in the average case, but fails in a relatively small percentage of case.

2) An algorithm which works in all cases, but is comparatively slow?

Assume the answer can't be the trivial "both, used when appropriately", because you don't have enough resources to implement both.

I think she makes a mistake in the problem statement - the issue with folksonomies isn't that the index is not "efficient" - it's that it improves average performance at the cost of non-average performance.

I don't know if there were more ping pongs between Dave and Elaine on this subject, but I think that the faster people like Elaine will recognize that Folksonomies are not here to completely replace Taxonomies the better.

I've also posted my 2 cents on the subject on my blog:
http://erans.blogspot.com/2006/11/folksonomies-taxonomies-and-coexistence.html

While it always seems to me that when people like Dave and Elaine argue its a clash of the Titans, even with my academic background, I still had to respond ;-)

Devil of outside, the content of the angel, return hesitant what? game

I highly recommend Thomas Vander Wal on Understanding Folksonomy at d.constuct 2006 - http://blip.tv/posts/stats/74688

I believe he coined the term Folksonomy?

In response to Dan's comment: as many spammers have noted (and consequently abused in the same way they previously abused meta-tags), the Google ranking algorithm is based on link structure, both internal and external to the document. Is this not a type of metadata, regardless of whether we think metadata has rigid types or not? ;)

Link abuse has a higher cost to the abuser (namely the cost of additional domains) compared to meta-tag abuse. I personally think the more metadata, the better. Ultimately, I think, that will create less possibilities for abuse, since an abuser will have to manipulate all metadata sources in order to be effective.

Well, David is right and Elaine is wrong, and that's that.

Yet, would Aristotle have been likely ever to wear a toga? Perhaps a tunic?

Here's a recent example of Folksonomy abuse.
Spam Hits Video Sites,
Misleading Web Viewers

"The simplest form of video spam occurs when users submit clips with misleading lists of keywords to describe them, such as using "Britney Spears" or "World Cup" to label videos that have nothing to do with the pop star or the soccer tournament. The spammers misrepresent their videos in the hope that the clips will be picked up in consumers' video searches, which generally rely on keywords."

nice post ! good to know what we shouldn't click.

agree with some of the arguments for strict classification and that there are problems with tagging en masse, though I agree with others that there is probably a critical mass where useless tags get washed out. I don’t agree, however, that using relative meaning is inheritently bad or invalid. In terms of discoverability and findability, all possible interpretations can be helpful. It can also provide context for works such as ones which had a political reason for being published even though the work itself was not about politics. I think the most useful benefit of tagging is the relationships between items that can be made apparent through the interpretations. I do agree, however, that there will also be the need for controlled vocabularies in organizations.

i think that tagging is an art....