Cypherpunks Goals: Bad debate drives out good debate

[Derek Upham]

Okay, I'll call the moderation method I suggested the ``asymptotic
moderation'' method.

Some problems that people have with asymptotic moderation:

1> It requires either a trusted subset of the membership to be
1> moderators and the most trusted are also the ones with the least
1> free time, or you trust everyone.  I'd hate to send a message to
1> Detweiler to see if he thinks it should be posted or not :)

We trust everyone.  Let's take a look at the four possible cases here:

Good message, good moderator: message accepted.
Bad message, good moderator: message denied.
Bad message, bad moderator: message denied.
Good message, bad moderator: SEND THE MESSAGE AGAIN.

If you're confident that your post was good, but rejected out-of-hand
by a bad moderator, just send the thing again.  It's highly unlikely
that the same bad moderator get it the next time around.

Obviously bad posts, though, will be rejected by everyone, no matter
how many times they get re-posted.  And just to be safe, the mailing
list program can keep track of rejection slips and refuse posts from
obvious abusers (and maybe it could do something similar with the
above denial-of-service attack).

2> The trouble with random single moderators are many, but worst would
2> probably be time-delay. If the chosen moderator for a message is
2> busy, sick, or away from their desk messages could be delayed for
2> days. The problem gets even worse if a delayed message is then
2> approved, and posted out of sequence.

The mailing list program can detect cases where people are taking too
long to reply; those posts can go to a designated list maintainer (or
maintainers) for immediate approvial.  People who are consistently
late will lose moderation privs.  Remember that posting out of
sequence isn't necessarily a thread-killer---Usenet survived for years
with UUCP transfer delays.

3> Instead of picking random list receivers to moderate, readers should
3> choose their own moderators.
3>
3> As a moderator reads the latest messages on the list, he or she can
3> mark each one as junk or not junk.  This causes advice messages to be
3> sent to their subscribers.  The subscribers can use mail programs which
3> process the advice and only show messages which have passed.  ("If all
3> three of my moderators say a message is junk, then don't read it,
3> otherwise, show me.")

So now we're back to kill files.  Here, though, the kill file rules
are based on out-of-band messages that can be received at any time
(possibly after you've read the message!).  And every user who wants
to use the kill file must set up killing software.  And there's no
guarantee that people would bother to rate every message they read
(I've participated in something similar on a BB, and _that_ particular
aspect failed miserably).

4> How about auto-moderation?  I came up with this idea a while back for
4> automatically moderating mailing lists.  Here's how it works:
4> [. . .]
4> If a person becomes a nuisance, people send their votes in to the
4> moderator-robot, and it tallies the votes. If within XXX days more
4> thumbs down votes are received than thumbs up votes, the person is
4> placed on the disapproved list.

Pretty good, but it wouldn't do anything to stop those people who
create random net addresses and post lone MAKE.MONEY.FAST or Jesus Is
Coming!!! messages.  And it's a bit too harsh to deal with
intermittent flame fests, especially flames from people who are
usually productive contributors (you know who you are).

5> There's a simpler solution.  Using the majordomo hack I posted
5> earlier, mail from known abusers would be bounced to the moderator
5> for his approval or rejection.  This would be a small volume, which
5> should be within the capabilities of one moderator.

How do we determine known abusers, and how can we deal with unknown
abusers or intermittent flame fests as above?


In summary, the aysmptotic moderation method has a couple of benefits.
All of the custom code is concentrated in one place; anyone with a
mail reader can perform moderation duties.  The moderation duties are
simple and well-defined.  You _know_ when you are being asked to make
a critical judgement (instead of judging everything---or more likely,
nothing).  All of the approval/disapproval information passes through
the central site, so that site can keep track of chronic abusers
through plain ol' numbers.  Finally, since the primary filtering
method does not depend on filtering specific users, it works just as
well for drive-by posters and for people who are suffering from a
temporary lack of control.


Derek

Derek Lynn Upham                               University of British Columbia
upham at cs.ubc.ca                                   Computer Science Department
=============================================================================
"Ha!  Your Leaping Tiger Kung Fu is no match for my Frightened Piglet Style!"