Technorati’s Kevin Marks gives an eloquent description of the advantage of free labeling, as opposed to predefined keywords, comparing Apple’s iPhoto keywords to Flickr’s tags1.
Being handed a list of keywords and asked to add your desired ones is in effect asking you to construct a personal ontology of the world; to break the world into categories you want to keep track of. To hold the entire universe in your head in one go, and chop it into meaningful chunks.
…
Conversely, Flickr prompts you for tags for each batch of photos you upload, and shows you each individual photo with a place to type tags in next to it. You look at the picture and type in the few words it makes you think of, move on to the next, and you’re done. The cognitive load is tiny, because you have the picture in front of you and you can’t help but think of words to describe it.
One thing we’ve been trying to do with the labeling feature in Confluence 2.0 is make adding labels as frictionless as possible. For labeling to work, you have to be able to add a label inline, while you’re viewing the content being labeled. You also have to give people the freedom to choose their own labels (so they’re not limited by some frozen, and usually wildly inappropriate ontology), but at the same time provide enough context in the form of suggested labels and type-ahead find that people feel safe they’re not picking the wrong one.
Of course, it’s helped to have a lot of prior art to learn from.
It all comes down to why wikis are successful in the first place, and why so many customers send us amazed emails about how readily Confluence has caught on in their business. Don’t impose structure up-front, because that will just discourage people from contributing. Allow people to create content as easily as possible, however they like, then give them the right tools to organise it and find it later. The structure of the site then evolves from the way people really want to use it.
1 Whether to call them ‘tags’ or ‘labels’ in Confluence was an ongoing debate for much of the early development of Confluence 2.0. In the end we decided that labels was the term that would be more readily accessible to the layperson, and anyway, you can’t go too wrong following Google’s lead, right? 🙂
