Wednesday, February 09, 2005

Tagging behavoir and quality

There have been, for a while now, a lot of data points and conversations about folksonomies swirling around in my head. Those have been added to by the nptech experiment, Rebecca Blood's recent comments about offensiveness and tagging, and a recent read at Darren Barefoot's site, Accentuating the Positive in Metadata and Folksonomies. I'm also thinking of my own use of tags which happens, primarily, at del.icio.us and flickr. I use tags to:

  • gather information from a wide variety of users;
  • organize material for my own benefit;
  • organize material for the benefit of a defined group of friends and family; and,
  • deliberately try to share things I think are interesting and/or useful with other, unknown, people who might be interested in the same topics.


What's lacking is any type of quality control.
The reason, I think, that these systems work so well is because they have figured out how to harness self-interest in the service of group benefits. So, quality control by ranking things (in the way that Stumbleupon does or even the Omidyar Network and eBay do) does not catch that particular tiger by the tail.

But what if tags functioned as a kind of vote?

That is, the number of times that something was tagged with a given word placed that something higher on the list.

Here's an example using flickr. Or anyway based on flickr.

Trying to invest users in a community while maintain the motivating factor of self-interest.
Right now, on flickr, you can set the ability of others to add tags to your photos. The default is that you allow your contacts to add tags. But this can be pulled in closer -- only contacts identified as friends and family can add tags -- or pushed out -- all flickr users can add tags. And, when tag gets added, I can see the list of current tags and, (presumably) it gets added for all users. That is, if I add "nptech" as a tag to an image, all users will find that image when looking at the "nptech" tag.

But what if that worked a little bit differently. What if the default is that any flickr user can add a tag to any public photo? And what if I can ignore tags that aren't mine?

For my own self-interest, I might tag other people's photos to be able to organize them myself. I might do this just because I want to, I might do it because I'm assembling collections of photos for various reasons. And I might want to ignore other tags because I don't want someone else's ad hoc taxonomy added into mine when I'm working with images within my own organization schemes. I might choose to this because I'm using a tag differently than other users, for example.

So, I'm tagging my own images and others. And when I'm tagging other images I'm not seeing the previous tags.

But this organizational aspect of flickr is only one of the benefits and completely ignores the social benefit. The ability to looks at everyone's tags and to tag surf images, to subscribe to tags etc.

So, when you move to everyone's tags, in fact, all of the added tags are there. But now, because users aren't looking at a list of tags and adding to it, but rather putting their own tags on an image, the tags have a natural weighting factor that is provided by the number of users that have chosen the same tag for an image. In many cases, that will be a single tag -- the tag provided by the images owner. So, now you can view the photos within a tag in two ways -- newest to oldest (that's how they are viewed now) or by weight. That is, if many users have used the particular word in tagging an item please show me that item first.

That provides a sort of quality control. I can tell that many people have felt a given word applied to an image. It also gives the community an opportunity to respond to what they may feel are inappropriate tags and they can do that by tagging an image with words they are feel are more appropriate. The image will fall to the bottom of the pile in one tag -- presumably the inappropriate one -- and will rise in another tag.

So, now tags are also a kind of vote. I can choose to view a collection of images -- including images that aren't my own -- in three ways: by my own organizational schema; by the organizational schema of all other users, newest first; and, by the organizational schema of all other users, weighted within a tag by the number of users who have chosen that particular tag.

It seems like -- seems like -- this continues to harness self-interest and doesn't interfere with the organizational or social aspects of these kinds of applications and begins to provide an opportunity for some sort of quality, editorial function via tag choices.

(in: tags, tagging, folksonomy, metadata, quality_control)