Why did you add it? What did you see that you didn’t expect?

Tags are more like noting properties than putting in buckets.

Originally built in text file with links, and starting putting a short text after the link, so he could ie. find all wifi links. That was the first use of tags. And then he built the web version of it, and it was still single user. Later decided to make it multiplayer. Interesting that there’s behavior of the group that has nothing to do with categories. Example is the “toread” tag.

Missed the question.

Some things have a clear category, for others it’s very nebulous. Contrasting the need for consensus in wikipedia with “everyone having their own space” in Contrast between people tagging photos for their own use for “ETech2005” when everyone else wants photos OF ETech 2005, and the pull between personal use and group use for something like Flickr.

Talk about the “circle in square” photos, tags which become a root for community and conversations. Another example was the glow project, pictures of stuff that’s glowing. Why the difference between groups and tags? Groups have permissions, different controls.

Canter asks about connecting tags together. How can we move tags around. Talk about exchanging database dumps at the minimum. That data is useful in many places. Joshua says 190K tags about, a lot of them single use. Provide tools to allow people to “trim the hedges”. As for shared tags, there are different axis of what you’re tagging, why, when it happens, who it’s for. In flickr you tag mostly for you, technorati tagging is mostly for others. Not all tags are the same thing. Saying they should all plug together doesn’t make too much sense, impedence mismatch.

A small group of domain experts can’t compete with a large group of self monitoring people when it comes to creating a category system. Paid people also act much different than volunteers.

How much metadata do we need about tags? Who created it, what tool created it, language, etc?

If you encumber them it doesn’t make sense. Tags are popular because they’re easy, have a low barrier to use. Flickr says that analysis is fine, but requiring more from the user is bad.