Default Settings

Why do tomorrow what can be put off until the day after?

I’ve been reading some classic texts on categorization and how the issue has been dealt with on the Web. User-generated tags, Clay Shirky writes, offer powerful, more “organic” alternatives to traditional forms of classification. Here I discuss tagging and some of the known problems with it, before relating the latter to a short case study on the ‘todo’ and ‘toread’ tags on del.icio.us.

Labels for Things

Shirky builds his argument around the concept of ontology - the question of “what kinds of things exist or can exist in the world”. Older classification systems, from the Dewey Decimal System to Yahoo’s Directory, offer a singular answer to this question. Things may exist within a pre-defined set of categories and sub-categories, and the world has been neatly accounted for. But there are a number of problems with this approach, including:

• Political bias - Having one group of experts present a singular vision is asking for problems. There is an element of power to classification, something that is most apparent in extreme cases such as racial classification in Apartheid South Africa (see the book by Geoffrey Bowker and Susan Leigh Star, Sorting Things Out).
• Instability - Things, and our knowledge of them, tend to change over time. Categorizations, however, often hold due to material constraints. (Rather than redistribute all books relating to the Soviet Union in 1992, the Library of Congress’ category was renamed ‘Former Soviet Union’.)

Tagging, on the other hand, allows users to add keywords to content they save, whether that is photos (in the case of flickr), videos (YouTube), or other content with a URL. It is a radically different manner of providing labels for things. In contrast to Web directories, it acknowledges the instability of categories and things, as tags can be adjusted over time. Tags are of course generated by users, who can act out of self-interest (bookmarking and labeling for easy retrieval). At the same time, though, del.icio.us and flickr can aggregate tags in ways that further benefit users - seeing all items tagged ‘wilfing‘, for example.

But more and more, we hear about problems with tagging. First off is malicious tagging, or the inevitability of spam. Fauxonomy is created with “metadata added with the conscious intent to confuse or obfuscate”. And there is another, more fundamental problem that Jan Simons pointed to recently. Ideas about what tags themselves are or should be - the over-rated ontology of tagging, I guess - can clash with how people actually use them. The definition of tags as single-word labels has impacted how we aggregate and make use of them (i.e. in ‘flat’ tag clouds). Simons has shown that tags used on flickr often fit within basic cognitive schemes (e.g. a photo will be tagged with a place, an event and an object). A question going forward is how this kind of research can expand and improve on the uses and presentation of tag-enabled classification.

Making Tagging Work / Taking the Work out of Tagging

But another option is to take measures toward making tagging work how it ’should’. In Protocol, Alexander Galloway argues that Web organization is, by necessity, radically inclusive and reactionary at the same time. To ensure that everyone tags, and that everything can be tagged, one would need tagging standards.

This was an issue for the collaborative blog I contribute to, Masters of Media, and it was interesting to hear different ideas on what kinds of tags should be used, and the various rationales for standards. Shirky writes that a major difference between old classification methods and tagging is that users are not ‘built in’ - del.icio.us users simply label out of their own interest, and consensus naturally emerges. In our case, however, we were debating how people make use of tags, what kinds of tags they look for, etc. In other words, we certainly were tagging for (imagined) users. And we were definitely not the first: a more significant list of tagging guidelines is found on Daily Kos, a popular U.S. politics community blog. (I would be interested to know whether these are still upheld by the community, but that would be quite some work.)

Where guidelines are unsuitable, offering suggested fields may also solve the problem. When I decide to save a page to del.icio.us, I get a set of recommended tags, based on what users have chosen previously. And when I begin to type, I get further suggestions based on tags I have used previously.

How do suggested fields affect individual tagging? As Shirky writes, the presence of difference, or multiple points of view, would appear to be a benefit rather than a negative consequence of tagging. Could suggested fields become tagging-degradation?

Todo and toread: what are del.icio.us users putting off?

One of the first tags to catch my attention on del.icio.us was the ‘toread’ label. Similarly to Simons’ critique, it is worth pointing out that this tag does not describe content, but a relationship between user and content. More precisely, along with its cousin ‘todo’, it refers to the deferral of an act. And in my own cynical reading, it refers to procrastination, and makes me wonder what exactly is being put off by del.icio.us users.

Using tools from the Digital Methods Initiative, I’ve done a quick analysis of the del.icio.us tags used in combination with ‘toread’ and ‘todo’. In other words, what kinds of content are being saved for later? The following are clouds of related tags from recent entries, and should give an indication.

In addition to the tags one would expect - i.e. ‘book’ and ‘article’ for toread - what I noticed straight away was the resonance with tags relating to productivity (e.g. lifehacks and lifehacker). I learned that one of these, ‘gtd’, means ‘Getting Things Done’, a central theme for productivity-guru David Allen. So what kind of conclusions can be drawn?

Put crudely, what del.icio.us users are putting off is becoming productive. This fits somewhat with Florian Cramer’s critique of folksonomic classification and the Semantic Web more generally:

[Tim Berner-Lee's] Semantic Web implies a complexity nightmare of meta information overtaking information, with each piece of information creating at least twice as much work for its semantic markup than for its creation proper, comparable to a library whose the catalogs outnumber the books they reference.

Add to this image one of the user restlessly tagging content, putting off its consumption, pre-filtering the Web for future use that never comes. Getting things tagged.

One could take this analysis a little further by finding the oldest uses of the ‘toread’ tag on del.icio.us. I did my best, but came across some limitations. (Individual tags, as far as I can tell, have not been attributed metadata such as timestamps.) Let us just assume that the user I looked at, who I won’t name, has returned to the 125th Anniversary edition of Science Magazine, tagged toread 3 years ago, or will someday soon.