Thursday 14 August 2008

Everything is miscellaneous - David Weinberger

Finally got round to reading this... some snippets below:

"Instead of everything having its place, it's better if things can get assigned multiple places simultaneously.

[...] we're no longer forced to carefully construct a single shared path through memory.

[...]the digital order ignores the paper order's requirement that labels be smaller than the things they're labelling.

knowledge is diverse, changing, imbued with the cultural values of the moment."

Weinberger argues that for the web to work, given the volume of information it now contains, that social tagging is the most feasible way to "organise" the content. But I'd argue that this assumes that all information is equal - and it's not.

Weinberger covers the different ways we've developed of organising information - from library catalogues and Dewey through to Amazon's "planned serendipity" and collaborative filtering (customers who bought this also bought x). He suggests that Amazon's offering of multiple routes (top-down categories, user generated lists, filtering etc) is an ideal combination of top-down and bottom-up.

How we organise information depends on the purpose of organising it in the first place - i.e. simply to map knowledge or to prioritise certain information above others. He mentions faceted classification "that dynamically constructs a browsable, branching tree that exactly meets your immediate needs". He also explores controlled vocabularies and thesauri - although controlled vocabularies can be too static, they can express relationships between terms. Weinberger points out the weaknesses of different forms of classification - either you are limited by the number of classification terms you can add or the terms chosen by the creator are not necessarily those used by the reader. There is also the question of how we name "things" - e.g. cancer is a broad term used to describe 100s of different diseases.

Weinberger argues that a system for organising information has to add value e.g. the periodic table. He suggests that bottom-up tagging adds value - in fact, the more tags there are, the more value. Tags can be shared and clustered - visualisation tools can help the user navigate the clusters.

The main thrust is that we can do more in the digital world than we could in the paper world. We're not constrained by the fact that physical objects can only be in one place at any one time. Digital objects can have multiple tags attached:
"How their [Delicious, Flickr, BBC, Wikipedia] content is actually arranged does not determine how that content can and will be arranged by their users [...] These examples are miscellaneous because users don't need to know the inner organisation, because that inner order doesn't result in a preferred order of use, and because users have wide flexibility to order the pieces as they want, even and especially in unanticipated ways."
To enable this "unanticipated use" what should identifiers point to? At the BBC, they've been exploring this - should the identifiers be at programme, series, episode, segment, frame level? Weinberger refers to IFLA's FRBR which has been designed to help with this problem.

Weinberger also gets into some detail about organising user-generated content, such as the reviews on Digg and Reddit. Some of these issues have been tackled by Wikipedia - although I guess the jury's out on how well they've succeeded in achieving accuracy (although it apparently measures up to Britannica). Wikipedia has developed a list of labels for non-neutral articles - kind of like a crude confidence interval, almost. He also looks at the explosion of mashups, with Google Maps inevitably getting a mention. User-generated content is growing at a phenomenal speed - Weinberger suggests that the more metadata (in the form of tags), the better we can manage this huge volume. He does mention the semantic web and the gulf between the vision and what can practically be achieved. The task now is not so much to build knowledge but to build meaning.

No comments: