Paul Gries and I were working at the same desk this morning. He was paging through dozens of posts to the bulletin board for a first-year course, many of which were repeats of questions that had already been answered. At the same time, I was setting up SpamBayes on my no-longer-new laptop. We looked at each other, and thought, "Why not use Bayesian filtering on posts to newsgroups and mailing lists?" If the filter decides that what you've asked to post is closer than it likes to something that's already there, it can point you at the previous message, and ask you to confirm that you actually want to post. Sure, people could do this on their own, but many don't.
Anyone ever seen anything like this?