I found the Babel Experiment described here very interesting. It claims to show that if you only allow people to use UML and not to talk, you get a better result than if you allow them to talk. The better result is only that "their diagrams were more
Read More...