Internetworkers: http://www.ibiblio.org/internetworkers/

["Michael D. Thomas" <mdthomas AT mindspring.com>]

> I did find it interesting that I was just listening to "The Connection"
> earlier in the week, which talked about just this sort of thing.
>
> "...picture the online equivalent of a barroom brawl..."

Interesting...

I'd like to propose the hypothesis that these online brawls often don't occur
in complete independence... In the same way that rain showers in Chapel Hill
and rain showers in Raleigh usually don't occur in complete independence.

What I'd like is an n-dimensional map of the Internet's flame wars and what
they are about. It could be modeled on blogdex, which has been tracking hot
new trends by examining thousands of weblogs. "Flamedex" would look at online
forums for the current areas of flaming conflict.

Blog wars would also be within the domain of the flamedex problem. The more
distributed nature of blog wars might make the implementation harder, but then
the lines of attack are often rigorously specified with URLs. Perhaps this
would make the shape of the battles (and thus the terrain) easier to see.

In both cases, you could get a sense of alliances and terrain. For instance,
you could look for "enemy of your enemy is your friend" symmetries. However, I
suspect that you'll find such symmetries to hold true only inside of  a
particular context. As you cross from one context to another -- i.e., when you
observe the behavior of a flamer in different and mutually exclusive online
forums -- I bet you'll find that their alliances are more emotionally and
tactically based rather than being ideologically based. After all, politics is
politics.

The hardest implementation problem is the same as with spam detection. As
someone on this list pointed out long ago, it is easy for us humans to
recognize spam, just as it is easy for us to recognize flame wars. But getting
a computer to recognize a flame war would be hard. Frequency of posts seems to
be a good signal. Also, as a flame war progresses, it seems that each message
becomes less and less relevant to the initial topic. This "loss of relevancy"
could also be a signal that could be detected with today's technology. I know
that Oracle Text can do thematic searches on text documents. The sales pitch
for that product is that you can do searches for documents about European
History rather than doing a search for documents that contain the words
"european history." To detect "loss of relevancy," you'd look for dramatic
changes in the themes of messages as a discussion progresses.

There's also a signal-to-noise ratio problem with flame detection. Who really
cares about the content of purely tactical and simple-minded flames such as
the spelling flame, the grammar flame, the etiquette flame and the dictionary
flame? Such flames are like petty assaults (and societal responses to such
assaults) that inevitably transpire during a larger and more important war of
words. The occurrence of such flames usually speaks of the aspects of a
particular online community rather than whatever ideas mingled with the
discussion that sparked the incidental flame.

Whadya think? Go ahead, flame away ;-)

Here are some links on the subject:

http://blogdex.media.mit.edu/

http://www.levity.com/markdery/flamewars.html

http://www.socsci.mcmaster.ca/soc/courses/soc4j3/stuweb/cyber9/flmg.htm

http://psychcentral.com/storm1.htm

http://technet.oracle.com/products/text/pdf/text_techwp.pdf