Tags are great, we all love tags! For one, they replaced the mentally taxing task of having to choose one and only one folder or category to put an item into. They also built upon the idea of folksonomy; that people over time will conform in terms of how they describe things because they both want to be understood and understand others. It’s an evolutionary process where the rules, albeit implicit, will be set by the users themselves and as such there’s no real need for restrictions or guidelines.
In some cases, like in gmail (labels) and in google reader (folders), users will probably use the tags like folders, because that’s how most of us are used to dealing with email and feeds. My guess is that most people will probably stick with this and many of them might even consider the possibility of assigning more than one folder/label to be more of a bug than a feature. In the case of gmail you’re not imposing your scheme on anyone but yourself but on the other hand, you’re not exposed to other people’s tagging either so you’re not conforming. Thus, the term folksonomy is not really applicable here. It’s first when you start sharing your tags that they become interesting.
So take Flickr, which employs what is called narrow folksonomy, where each item has only one set of tags associated with it, whoever adds a tag to an item, adds it for everyone else as well. Here I think is where we’re currently seeing folksonomy at its finest, where it actually works… To a certain degree. Because, when you want to search, do you really choose to do that by tags alone? Or when flickr only had search by tags, did you actually find everything that you wanted that way? Or when you want to find anything from your archives, is the tag cloud your first association?
There’s also broad folksonomy tagging, as employed by del.icio.us amongst others. Here each item is described by each user individually. Depending on if you look at it from an individual user’s point of view or from the items’, you get two different but overlapping tag sets. One is a union of all users’ descriptions of that item and the other is a subset of the former. The interesting part is that with del.icio.us, it actively tells you how to conform in terms of tagging folksonomy. That is, based on how other people have tagged the item you’re adding, del.icio.us is giving you suggestions on how to tag your item.
Great! So now you have all your items properly tagged, conforming with how other people has chosen to tag the same items. Only thing is, they’re all tagged with web2.0 and social! Not very surprising since those were the most popular choices and they were displayed with the biggest font. They were also the most general of the choices and thus wholly seemed to describe the items. Only not precisely so.
I think one problem is that we’re not all biologists. Nor economists, nor linguists. A folksonomy, if employed by a larger population, will conform towards the generic, not the specific. To a flower, if you will, and not a rose. In the fields where there’s already an established taxonomy, folksonomy will water it out. In the fields where a clear taxonomy not yet exists, such as in web applications, folksonomy through tags as they are employed today will not create one. After all, most people aren’t really setting out to be taxonomists, they just want to add an item to del.icio.us and be done with it. If someone says it’s web2.0, who am I to argue—who even cares?
However, tags came out of the need for something that was a bit more free than categories and folders, yet still more strict and powerful than plain text. In actuality, tags also came out of the real-world problem that freetext search in big and constantly updated databases was not feasible. Powerful as tags were meant to be for both the generic and specific, they were the most powerful when it came to drawing the broad strokes: to show trends and what is currently popular. Pair a very general tag with a current event or product release and suddenly it becomes crystal clear what it means.
In a sense, applications that use tags based on their popularity are doing their users a disservice. Because it really only aids them when they are adding an item, not when they’re later trying to find it. And if they’re not storing things in order to find them later in the first place, there’s no basis for folksonomy and shared tags could be skipped altogether.
What can be done though? In theory, perhaps the following:
- Separate the broad from the specific.
- Aim for the benefit of the long tail. If you can’t get them to work for that, your shared tags are probably not working at all.
- Separate search from tags. They should not just be a subset of search.
- Exclude popular tags and those shared by many users.
- Keep filtering until your tag cloud is both intriguing and surprising.
In practice, it’s not going to be that simple. However, that’s where the creative people in your staff come in, go put them to good use, give us tags 2007!
20070319: What about a bell type curve?
