You and me both. My thoughts on this subject are that all knowledge is related in some manner, so unless you want your knowledge points to form an all to all connection (useless), there needs to be some probabilistic connection.

This means that the user blanket accepts the probability that language is similar enough in two separate thoughts to form a connection, or they receive a basket of suggestions from a program and approve/disapprove each. One inevitably leads to meaningless connections, and the other leads to work on the user's part to prune out all the trash.

What do you think?