On Folksonomy, Feedback, and Polysemy
There has been some debate and discussion over the buzz word “folksonomy.” A quick definition is that a folksonomy is an organizational system reliant on multiple individuals applying personalized tags (labeling terms) to the categorized item. These tags, as more and more of them are added, are then used as the organizing elements for the system of objects. Thomas Vander Wal has given the best explanation I've been able to find, including differentiating between broad and narrow folksonomies (ex: del.icio.us and Flickr respectively). Such a system encourages distributed relationships between specific words, objects, and individuals. Functionally this network is decentralized, with each node—be it a tag, individual, or object—able to connect to another within the system. What arises, when enacted on a large enough scale, is rhizomatic: multiple points of entry for multiple participants—via one of the three main types of node—which expand, ideally (that is, if individuals tag objects in a highly personalized manner as Vander Wal seems to require), without interference from a central or centralizing entity. The primary advantage to such a system is that categorization becomes malleable to an individual or group's specific vocabulary, developing greater ability for one node to connect to another.
Where Vander Wal's description falls short is the assumption that individuals have the ability to tag without influence from the tagging system itself. When tagging a new item on del.icio.us, for instance, one sees suggested and popular tags. Furthermore, a key component of a folksonomy-based network, such as del.icio.us, is social interaction; users will inevitably be influenced by the choices made by one another. Vander Wal wants to exclude this type of influence from the tagging process under the assumption that it would distort the potential emergent tags (about emergent tags: 1 2). This systemic feedback, however, is unavoidable and unproblematic for a folksonomy. Emergent systems are reliant on both individual interaction and feedback from the larger system to the individuals. The fact that del.icio.us User A decides to use similar tags as User B or a “popular” tag does not make the contribution of User A to the system any less valuable. User A's individual interaction is dependent on the intersection of the multiple environments surrounding his tagging decisions, making it impossible to isolate a single environment as adversely interfering with the process. Any tags used will inevitably be a result of a combination of surrounding interferences.
The Wikipedia article on folksonomy notes polysemy as a problem with folksonomies. This is an understandable immediate reaction: a tag with multiple meanings will lead to multiple, thus on some level inaccurate, results. Yet rather than polysemy posing a problem for a folksonomy, I would suggest that folksonomy poses an insight into polysemy. A developed folksonomy would have many objects linked to many different tags by many individuals. To see the objects linked to a particular polyseme, as well as concurrent tags used by individuals, would not only provide a demonstration of dominant meanings but also the relationship between the different meanings of the original tag and other tags. What arises from this is a mapping of the separate networks surrounding, individually, the multiple meanings of the polyseme.
I have not tested this last suggestion. Results and methods of any such test will be posted here.
Comments
The polysemy problem is with tagging and not with folksonomy (if Wikipedia had a proper definition of folksonomy it would be clear). The folksonomy actually provides a solutions to the problem as people are less likely to reuse the same tag for differing items than the whole of a community. But when the individual does use the tag for more than one definition the other tags they apply to the object normally make it very easy to discern the tag's definition.
This is one major value to people using their own terms in tags, rather than relying on what others call them. But the main reason I find it important for people to tag with terms in their own vocabulary is for their own refindability of the object. When people use the terms of others the refindability drops off quite a bit as it is not the terminology that is most used by them and not familiar enough to use it as a search term for their own items.
The emergent values are also essential, but the prime value is for the tools to work as a tool for recalling their own objects of interest. There has to be inherent value for the people themselves using any social tool for them to continue use of it.
Posted by: vanderwal | January 8, 2006 10:49 PM
FYI, all the links in this post are broken so you might want to reedit it to fix them.
Posted by: lia | January 11, 2006 03:14 AM
Hmm, odd, thanks for letting me know.
Posted by: nathan | January 11, 2006 10:40 AM
Thomas, I believe you're making an assumption about the type of apps that use folksonomies (where folksonomies grow?). Yes, Delicious.com was designed primarily to aid individual memory, but it's entirely possible that other apps might be designed to make online resources more findable by others. E.g., a knowledge management well might want to add tagging primarily as a way of making individual discoveries available to an entire organization. In such a case, the folksonomy that organization needs is helped by showing taggers the popular tags as they are in the act of tagging precisely so they won't fall into the "trap" of using tags that are meaningful and memorable to them but not to others.
It seems to me that either type of tagging system can be useful, depending, of course, on the aims of its users.
Posted by: David Weinberger | January 13, 2006 03:51 PM