I've put a screen capture of a visualization of the Brin-l social network
(2001-2002) here:

http://www.mccmedia.com/brin-l-network.jpg

This is a 3D network, so some features are not perceptible in this static
shot.  However, I think I rotated it to a position where the major features
are visible.  The only significant illusion is that Darryl Shannon, John
Giorgis and Julia Thompson in the lower cluster appear to be closer together
than they really are.  But they are still fairly close when it is rotated
X-Y.  The upper cluster is tight no matter how you view it.

The source data is about 3200 name pairs that came from people who started
threads and the first two people who responded to the thread, subject to the
kind of errors I mentioned previously.  The people displayed are those with
20 or more links between them.  That thinned out the population enough to be
able to make out individual names.

You'll notice that people appear twice.  They appear once as an initiator of
a thread and again as a responder.  You may not see everyone's name twice,
though -- the visualization is zoomed in enough that outliers are off the
edges.  The upper cluster is mostly responses; the lower one is mostly
initiations.

Line thickness is an indication of the frequency with which people were
linked to one another.  You'll see a big line between Ilana and Julia, which
is due to the aforementioned threading error.

This is getting interesting enough that other work permitting, I'm going to
use my own archives to get the threading right, looking at the body of the
messages for confirmation.  The main reason I don't do that for other
communities is that it is very time-consuming to retrieve all the message
bodies with a web robot.

Nick Arnett
Phone/fax: 408-904-7198

Reply via email to