Northern Voice – Stephen Downes

Cindy and I attended the Northern Voice blogging conference today. We thoroughly enjoyed it: the setting was good, the weather was beautiful, and the whole thing was a comfortable size. Many of the speakers were excellent. I intend to write several up, but I’ll start with Stephen Downes' presentation on Community Blogging. He had much to say, so this is a long post.

Stephen made a couple of controversial claims, which dominated the question and answer. He argued that folksonomies are bad because they promote the long tail, which he also criticized. I think his opponents missed the real point of his presentation, which was must much more interesting.

RDF and Peer-to-Peer Search

Stephen emphasized that the meaning of a piece of information is not all inherent: much of it is a product of context. So the meaning of a blog post is not just the words of that post, but the links to and from that post, how those words are used elsewhere in the blogosphere, the other work of the author, etc.

He criticized top-down approaches to meaning, such as hierarchical systems of categories or centralized social software. Instead, he argued, meaning is produced through diversities of context: meaning should be produced not just by authors but by readers – what he called “3rd party metadata”. Meaning emerges from the community rather than defining the community (as it would in a taxonomy).

Stephen further suggested search based on relationships: rather than arbitrarily matching a tag from a taxonomy or folksonomy, or searching for keywords in Google, a search would spread out over the network, following links to related resources and using the structure of the system itself to collect meaning.

This fragmented (I am tempted to be impudent and say postmodern) meaning is what RDF is all about. It records statements about resources: statements that need not be coherent, complete, or centrally managed. This contrasts with the unified model of XML, which mirrors the record-oriented structure of most programming languages and databases. And such a search would naturally be peer-to-peer, relying almost entirely on context. This in turn cries out for a system of authority and identity; Stephen implied as much when he talked about using indirect relationships by author to construct and determine meaning.

Paradoxically, most identity schemes are centralized. This pursuit of diversity and context may trade centralized meaning for centralized identity.

The Long Tail

Stephen attacked the long tail for its power imbalance between the A-list bloggers and everyone else. Popular media sources – A-list blogs, for example – seem to control the agenda: they’re tremendously widely-read, popular, and influential compared with everyone else. They are in control.

My understanding is different. The idea of the long tail is that this is an illusion. In aggregate, all of the other blogs with their marginal readership and their micro communities are actually more powerful and influential than the A-list. Turn the power curve on its side and this it becomes obvious: the long tail trails up indefinitely, while the A-list on its side ends abruptly. Until the Internet, there was no long tail: after the A-list, the curve simply terminated. The long tail is not equality, or even close to it, but it offers hope.


Stephen’s criticism of folksonomies was that they are so general as to be useless; he used Technorati’s list of the top tags as an example: one of the top tags is “general”. But, as I have argued before, readers using networked folksonomies – as in – can slice a topic in multiple ways; with proper analysis, this creates the potential for very rich metadata.

Stephen also argued that the most popular tags encourage authors to over-use them, only amplifying the power disparities of the long tail. This reminds me of the communication theory of the “strength of weak ties”:

human communication typically entails a balance between similarity and dissimilarity, between familiarity and novelty . . . a new idea is communicated to a large number of individuals, and traverses a greater social distance, when passed through somewhat heterophilous links (as in radial personal networks) rather than through homophilous links (as in interlocking personal networks) (Rogers & Kincaid1, Communication Networks, 1981, 128).

In other words, exchange of information between two people requires on the one hand that they share enough in common to be able to communicate (they are homophilous), and on the other hand that they are different enough that they have something new to say (they are heterophilous).

If a tag is overused, it becomes useless: it cannot be used to discriminate. If it is underused, it is invisible. Stephen gave the example of a search for “physics photos” which turns up many entries marked “physics” and one marked “rabbit”. It seems to me that in this undifferentiated collection of physics content, it is the rabbit that stands out and provides really useful information.


1 Update 2005-02-25: I had neglected to credit Kincaid.