Made out of people

Categorization is the New Folders

Notes: Categorization is the New Foldering

Notes based loosely on the session from FOO Camp. Thanks to Ross Mayfield, the guys from deli.cio.us, Matthew MacLaurin and a bunch of other people who were nice enough to say their bit.

By "categorization," I mean specifically metadata that assigns labels and tags to individual items: files in file systems, for example, or email messages. The tags I'm interested in are not mutually exclusive--a single item can fall into more than one category--and are not (necessarily) hierarchial.

This is the philosophy behind GMAIL, DELICIOUS, FLICKR, and other tagging systems. It's the philosophy that was tried with earlier versions of Lotus Notes (which then managed to use a confusing "folder" metaphor).

Issues:
* Where does the metadata come from? When is it assigned? How easily can it be changed? Should categories be chosen in advance (and thus often turn out to be inaccurate) or should they evolve?

** If they evolve, how do you reconcile newer versions of where stuff goes with earlier versions? What happens to our deepening understanding of what the relevance of entries? (This is the autocollaboration to which I referrred earlier).

• Are the categories trying to be an exhaustive coverage of possible topics, or is the user comfortable with lots of stuff in "other" and overly-broad categories?

• Is the categorization created by individuals, or is it collectively derived? If the former, is there ever a need to reconcile with others? If it is collective, how is it arrived at?

• How does one handle linking categorization to the various namespaces available out there? Can we distinguish between (for example) all copies of a given book (Amazon's ASIN, or the ISBN) as opposed to a particular copy of a book? Can we point to an individual's files? (In some sense, this is a realization that while library systems--like LoC and Dewey--both identify something and tell us about it, this loose notion of categorization may not be able to do both.

• How can social navigation and related concepts be used to understand which categories a user is creating well (that is, successfully), or badly; how their categorization might match others (by pivoting on the object, for example); by statistically deriving similiarities.

September 13, 2004 08:28 PM | TrackBack | in Data and Documents