Seeing the wood for the trees

Caught up with an interesting two and a half year old piece by David Weinberger today: Trees and tags - Introduction. As so often, Weinberger's creative - almost poetic - use of analogies brings into focus a topic that had only been in my peripheral vision up to now. This time it was social bookmarking.

I've been using del.icio.us for about six weeks, mainly as a practical, personal solution to an out-of-control collection of miscellaneous bookmarks that I never quite got round to tidying into nice sensible folders. It works, on that level, & I've happily integrated it into my Facebook profile, & even been aware (in a peripheral vision sort of way) that this was Doing Web 2.0.

Weinberger puts the significance of social bookmarking like this: "autumn has come to the forest of knowledge". He doesn't mean that information is gracefully declining into old age, but that the old models of knowledge as tree-like structures are giving way to "piles of leaves". The old way of categorising knowledge is linear, static, inflexible: information lives in one place only. If I want to find out about Paris, for instance, I follow a path through Geography, to Europe, to France & finally Paris. That's how libraries of physical books are organised, & how the first web directories worked. It's what Weinberger calls a "traditional taxonomic tree". The problem with this way of organising knowledge is that it requires prior familiarity with the subject. It's like the dictionary paradox, that you have to know how a word is spelled before you can look it up to find its spelling.

A newer way involves "faceted classification" where there are several possible routes to a piece of information: I could reach Paris by choosing between places/people/activities in three different ways - 'places' most closely resembles the previous route, but 'people' would also work if I was looking for famous Parisians & 'activities' would work if I was planning a holiday. This is the classification system used by many e-commerce websites, where users can locate a product by drilling down through various options - price, brand, features & so on. It still involves a branching, tree-like hierarchy of information - a taxonomy.

Social bookmarking, by contrast, implies a model of knowledge that doesn't start from the trunk or the branches, but from the individual leaves. Suppose I type 'Paris fountains' into a search engine & find this website. I can then add it to my del.icio.us collection & tag (label) it as "travel" & "Paris". Any other del.icio.us user looking for information on Paris will find that site & by searching for "Paris" I will find what other users have tagged. My own Paris collection includes a blog with daily photos of the city, which is also tagged by other people as "blog" or "photo"/"photography". If this was a book, somebody would have to decide if it belonged on the Paris shelf or the photography shelf. In a non-hierarchical classification system, it can live on both.

What's more, the classification process is entirely democratic. No experts need to judge whether each website is worthy, or how it ought to be categorised. Users themselves decide. We can kick aside the pile of leaves, or toss it in the air, or pick out individual leaves & stick them on paper to make a picture, or even hang them on a traditional tree of knowledge. The decisions we make will in turn influence other users. These are no longer taxonomies but folksonomies: knowledge systems organised by "folk" - you, me, ordinary people. As Weinberger puts it, tags & social bookmarking are "an inexpensive, easy way of using the wisdom of the crowd to make resources visible and sortable".

Just as an aside, it struck me that that the TV show "Who wants to be a millionaire?" draws on the three orders of knowledge summarised by Weinbberger. Contestants have three 'lifelines' to assist them in choosing the correct answers to quiz questions. One is "phone a friend" - this amounts to consulting an expert, so fits well with the professionally controlled tree model of information. Then there's "50-50" where half the possible answers are eliminated, to increase the contestant's chances of choosing the correct one. This approximates closely to the faceted model, where users choose between successive options to narrow down their route to the right information destination. Finally, there's "ask the audience" where the studio guests are asked to vote on what they believe the right answer to be, to guide the contestant's response. This, surely, is Web 2.0, the "wisdom of the crowd".

This one doesn't exist yet

I like Martin Weller's suggestion of a meta web 2.0 service ("We do web 2.0 so you don't have to")

what I need is a meta web 2.0 service, one which finds the web 2.o services I need because I can't keep up with them all. I can't even be bothered to work out which fine niche each service is targeting, and whether that is in fact a service I need.

I have a long list of web 2.0 applications to explore, but there aren't enough hours in a day to go on safari, let alone send back deep & insightful blog posts from the field.

There's also a faint echo of some half-learned economics - the distinction between needs & wants, & capitalism's survival on the back of creating wants we didn't know we wanted & needs we didn't know we needed - making me suspicious of web 2.0's proliferation. Perhaps this is all pascalian diversion from What Really Matters.....

Even so, Martin's startup proposal might be the solution. Would it be a 'Rough Guide to Web 2.0' or something far more sophisticated & personalised? Would I be prepared to accept the degree of self-exposure it requires to function usefully? As Martin explains, such a service could only operate if it was primed with detailed knowledge of my preferences, routines, habits, like a sort of online PA.

Given the rather crude attempts of Amazon to tell me what kind of books & CDs my previous browsing/purchasing history suggests I would like to "consider", or the OU Courses & Qualifications website's hopeful listings of courses that 'students who studied this course have also studied at some time' I'm a little dubious about how much information I would need to provide, for a meta Web 2.0 service to achieve greater efficiency than I could myself.

I'm also curious as to how it would search the available options, which change daily. I don't think automated searching could come anywhere near capturing what each service offers, unless there was an agreed set of searchable metadata available (don't get me started on metadata: I encountered them once before & came off worse).

Despite these reservations, I still think it has promise, if only as an expression of our sense of helplessness watching something unfold that we cannot hope to grasp in its entirety.