I was rereading this post from 2008 this morning and thinking again about this little model, spotted then by Roland Harwood. The gist of the post, and Roland’s argument, was that we tend to overvalue networks that look like A and B, partly because we can’t follow all the conversations