Sheetrock's Journal: Neomoderation

Please feel free to comment on CmdrTaco's latest journal entry here. Sadly, there is no location in the entry itself to offer insight into a new moderation system or helpful office moving techniques, which creates a dearth of information and dialogue regarding both subjects. Please start the header of any goatse related comments with [goatse] for easier filtering and sorting.

Moving-related tips:

• Lift with your knees

Now, moderation is something that we could have a 600 comment discussion on... if it wasn't relegated to my pathetic corner of the website. Easily finding the good parts of said discussion might be a bit tougher for obvious reasons, but it'd make a good Ask Slashdot topic. Not that I'm about to submit it, but you can.

Remaining aware of the constraints of keeping the system responsive, keeping moderation simple, and keeping the selection/filtering of comments to read easy for the user, I think the concepts raised in the journal entry are good ones. For example, maybe the user could use better granularity in highlighting/sinking messages with particular content, such as being able to bump messages up that were 20% or more Funny by moderation. Maybe the moderation categories could be a little more helpful and moderators could rate a comment on more than one parameter: +1 Well-written, -1 Flamebait, -1 Incorrect on the same comment, for example.

Bayesian filtering in the way in which I have been accustomed to using it on spam sounds expensive, and infeasible on a per-user basis, but I assume he's talking about using it either on a subscriber-only basis or over every comment entered into the system in conjunction with the score of the comment to subtly affect the score on other comments with similar content. Or in some other way completely different from what I'm thinking.

I think it'd be interesting to rate comments similarly to the tipjar model. Everybody gets twenty points in the beginning of the week and is allowed to spend as much as they want on any and all comments (maybe even their own?) Promotion only, no downrating, no ceiling on comment scores. Or perhaps stagger the handouts, giving different sets of users twenty points every day such that by the end of the week everybody's had twenty. Obvious problems with multiple accounts, but perhaps those could be worked around or mitigated.

Then again, everybody's got a moderation model they'd like to see, and barring some kind of simulation there's only one way to test it (hopefully in parallel with slashdot.org, as was done with brak.slashdot.org). But there are other features that might improve the ability of getting the most meaningful information in front of the reader quickly.

Being able to see the newest comments in a story (hopefully fit in with the user's preferred reading method -- nested, threaded, flat, etc.) would not only save nested browsers scanning time but also much bandwidth. Well, one can work some of this functionality with 'Flat' 'Newest First (Ignore Threads)', but it'd be helpful to have a mode where older comments are threaded and newer comments are nested (fully visible) beneath them. And maybe very complex to implement too; I don't know.

Any thoughts?

Edit: Also, smileys??

new/old threads

[croddy]

this is actually fairly easy to implement. I have something similar going on my php-based webboard.

the trick is to create a new database table, and store userid, story id, and lastpostdate and/or highest comment id in the story at last read. we just zap it forward to an html anchor though, but doing a second query wouldn't be too difficult.

then again, my board's not written in perl :-P

Similar to "tip jar"

[Jouster]

Each (week||day||month||hour||.*), each user gets a number of points equal to their karma. When reading a discussion, you have the option of how many points to assign to an entry (-5 to +5), but it's not actually assigning that value; that's just a proportion of your points that you're adding. For instance, if I had fourteen points, and marked one comment +5 and another +2, it'd give 10 points to the first and 4 points to the second.

M2 takes a much bigger role in this sort of system, especially early on,

Another one...

[Jouster]

A different approach would be to make everyone moderators, with complete authority (including "bitchslap" of threads!). Logged-in users choose their moderators; this is the perfect way to accomodate different tastes in humor, for instance. Several people might focus on spending their mod points on humorous posts, and you just choose the one you think is most in line with your tastes.

Subscribers get the ability to assign relative weights to their moderators.

Friends lists are the default moderators lists,

Re:Another one...

[Jouster]

Oh, and in case anyone's worried about how much it costs (CPU-wise) to do a "web of trust" in real time, and potentials for googlebombing the mod system--the analysis would be done each night, and in the morning the site administrators would evaluate the produced list in order to choose who acts as the non-logged-in moderators for the day. Anyone who called it censorship would be told to create their own account and subscribe to other moderators; the editors would merely be making sure some huge script to

Re:Another one...

[Sheetrock]

It's interesting you mention this. I was (idly, not actively) speculating about the concept of grafting a new moderation system over Slashdot, as follows:

Everybody interested in participating in such a system could run a client that acts as a proxy. This client fetches a Slashdot page and alters it slightly such that links to other Slashdot items go through the proxy. When you view a story, a regexp (love those things) adds a moderation option in each comment header and a button that directs the client

There is another way to moderate.

[t]

If the comments could be downloaded like a newsgroup then people could start tinkering with everything from bayesian filtering to some type of natural language processing. Sadly that won't ever happen.

As it sits now you have a small group of already very busy people who lack any expertise or even education in language oriented algorithms versus a possibly large and diverse group of trolls who can try anything they like, as often as they like. There is just no way the /. crew can win so long as the game is

I resemble that remark.

[Sheetrock]

As a user that greatly enjoys discussion about the potential merits and pitfalls of Intelligent Design theory, the dangers posed to society by creating and disseminating 'murder simulators' in the guise of video games to our youth, and the true introduction of chopsticks to the Asian mainland via immigrants to American mining communities in the 1800s, I am perhaps not the natural choice to champion more stringent moderation methodologies. So please do take my thoughts with a grain of salt.

It may be possi