From: "Vladimir Z. Nuri" <vznuri@netcom.com>
Date: Tue, 14 Jan 1997 08:52:47 -0800 (PST)
To: cypherpunks@toad.com
Subject: s/n problems on this list
Message-ID: <199701141652.IAA10713@netcom13.netcom.com>
MIME-Version: 1.0
Content-Type: text/plain


I've posted on this subject many times in the past, and its
something I like to brainstorm on. signal to noise problems
on this list. 

generally, the moderator-specific approaches I find less
palatable than techniques that can be automated. here
are some more ideas for automated moderation techniques.

1. software called "grouplens" I've talked about here (search
yahoo) has been used to rate articles. they found that there
was a very high correlation between how long people spent reading
articles and the "interesting" rating that they gave that article.

hence, a proposal: cpunks have long advocated and hacked mailer
programs to stick in crypto. how about sticking in a hack that
tracks how long mail messages are being read, and send that
info back to an auxilliary mailing list address. the list
software keeps track of time spent reading articles and
can allow people to screen the list based on the "most interesting
articles". "send me only those articles that go over a threshhold
of 500 combined human attention minutes".

of course this can be messed up by hackers who try to skew the
ratings. I would suggest a limit on the max time per article per
person.

2. I like the idea of a system that keeps track of complaints against
given users. it could keep track of the complaint/per age of 
every subscriber. e.g. hypothetical: LD gets 10 complaints per week, and TCM
gets 30 per week on average, and it could allow people to screen
the mail based on these numbers, such as a command "send me mail
of all users below a 20 complaint/per week value". you could,
for either of these approaches, have an "official screening value"
in which the owner of the list adjusts these parameters to his
liking.

just more ideas. I like cpunks generally favor a technological
fix that doesn't involve bottlenecks of individual moderator
opinions. (2) would be pretty easy to implement inside existing
mailing list software, at the majordomo side.