Collaboration Is the Database

When I worked at the former Institute for Learning Sciences at Northwestern University we used to create things very much like Wikis. I worked with a kind of hypermedia style innovated, but since not very popular, called an Ask System. An Ask System would view a long conversation as a cluster of answers to questions asked.

The basic unit of a ask system was a clip, generally a story or short answer lasting no more than a minute. A story could be seen as generally answering one or a few related questions. We would watch long videos, break them in to clips, and then connect the questions the story answered to the clip. Then we could further add the questions a clip asked and the indexing would be done by linking questions asked by a clip to other clips that answered that question.

The main take away I got from this work is that it is very hard, if not impossible to formally capture all the reasonable networks that can exist in a corpus of data. This is the "Wiki problem" as I see it. The number of links that could be made keeps growing as you add more and more clips until soon your system is noteworthy for the massive supply of unmade links.

Following links also can easily take you down blind allies. It was my experience with Ask Systems, and also with Wikipedia since, it was good for exploratory context learning. But it was not as good for finding context specific knowledge. Ask Systems had an advantage over Wikipedia in that knowledge was represented by the questions it can answer, rather than just tags. Tags are fine for dictionary like definitions of knowledge but how useful are these. It is very rare I get a question like "What is this term mean" at work. Most likely I get questions like "how can I do this?" "why would I have this problem?" "what do we need to do this?" Planning information requested in the form of highly contextual questions, where the real value added comes in.