« Adventures in Agile : Part 2 | Main | Iteration 6 signed off »

December 20, 2005

Custard Ontology

The custard project is the culmination of an awful lot of thinking. Like most social software projects it owes a lot to the Cluetrain Manifesto, and my contribution was probably influenced by stuff like Gonzo Marketing, Permission Marketing and the rest of the usual suspects. Somewhere at the back of every NGO web project is the ghost of the Dean Campaign, promising endless dollars if only we'll do whatever it was they did. But there is one piece of external thinking that has probably done more to shape this project than any other, and it's this article - Ontology is Over-rated - by Clay Shirky.

The Clay Shirky article is interesting because it goes to what has become the heart of the project. Ideas about tagging and classification. Shirky argues that rigid categorisations are difficult to impose and often lose their value in the face of a changing world, besides which they're fundamentally rooted in the idea of arranging physical books on physical shelves.

On the internet, he says, there is no shelf. To prove it he's got a nice series of diagrams... (go and look, then come back)

The one we're most interested in is that last one. Which I'll reproduce here to make life easier.

[Just links (There is no filesystem)]


Here the rigid hierarchical structure of the earlier system has been replaced with a structure composed entirely of links between different pieces of information. The links form an unlimited number of possible shelves and filing systems, and by allowing anyone who wants to make connections between disparate bits of data the system grows and structures itself.

This is why services like del.ico.us and Flikr work, and hopefully it's why Custard will work. Within custard there are three ways to link pieces of information.

1. Tagging allows for one to many relationships. A tag can be shared by any number of pieces of content. 'Power' might be associated with pictures of a power plant, and the President of the USA. When tags become used in combination more clarity emerges 'Coal Power' shows the power plant but not the president, what 'Oil Power' shows is a different matter, and without making a value judgement about the content which we expect our users to share, who are we to tell whoever made the connection that they are wrong?

2. Linking allows for one to one relationships. It's a way to express connections between pieces of information that go beyond a shared subject. For instance you might link an event, to a map (showing where it is), a discussion group (who organised it) and a press report (describing what happened). These items will each have a place of their own among the multiple shelves of the tagging system, but once a visitor has found a way to one, accessing the others should be simple.

3. Geo Location where a thing is, or where a thing is about is a good way to connect things. Pollution near powerplants is one example, activists near other activists is another. Custard will make it easy to make these connections between different pieces of content, providing another dimension for our increasingly flexible shelving system.

At the end of the current iteration (Thursday, due to some folks being ill / on holiday / whatever) we'll be able to show the world a custard where linking, tagging and geo-location are all running together. We expect that this will lead to some changes about the way we see the system and interact with it. With luck it will bring to mind some new possibilities, and hopefully show the rest of the world what it is we've been working on for all this time.

Posted by Martin Lloyd at December 20, 2005 1:18 PM

Trackback Pings

TrackBack URL for this entry:
http://weblog.greenpeace.org/cgi-bin/mv/mt-tb.cgi/892

Comments

Martin, you might like to read http://www.aripaparo.com/archive/001456.html

Posted by: Thijs van der Vossen at December 20, 2005 10:05 PM

Yup, that's interesting. It's also good to see that del.icio.us has been able to achieve what it did with a pretty small group of users. 300k users is not a lot, especially when you assume that a lot of them will have signed up once to see what it was all about and then gone away.

One of my concerns about this project has been about what establishes 'critical mass', the point at which there's enough value in the tags to encourage people to exploit them and add more. This suggests the threshold might be pretty low, especially if we start with small 'microcommunities' where there'll be a lot of similar content grouped by a limited set of tags.

Posted by: Martin at December 22, 2005 9:59 AM

Post a comment





Remember Me?