Re: Moderation at scale is hard^Wimpossible, or Why we really can't use Facebook for technical discussion.

[John Levine <[e-mail]>]

It appears that Phillip Hallam-Baker <[e-mail]> said:
>Slashdot has a rather less respectful audience yet their karma system
>produces vastly better results. But the biggest takeaway for me is that one
>size fits all does not work for curation or moderation.

Sure, but differences in scale have qualitatitive as well as quantitative effects.

Facebook has 2 billion users who send about a million messages a
second. Slashdot has 4 million users who send about a three messages
per minute. (These are numbers from their web sites. Slashdot also
owns the much larger Sourceforge which I'm not counting here.)

FB is big enough that there are entire businesses and political
campaigns that live on it, with large financial consequences depending
on who sees what. The stakes are much higher, and the scale is so huge
that no matter how large their farm of human moderators, the humans
will never see more than a miniscule slice of the traffic.

Large platforms know that their moderation software sometimes guesses
wrong, and they deliberately make it fail safe -- compare the effects
of taking down your snarky message or Dean's vs. leaving up actual
death threats.  Is that really surprising?  Perhaps the message here is
that Facebook is not the ideal venue for nuanced political debate.
Once again, on a free service, you often get what you pay for.

For some interesting insights into moderation at scale, this blog
post about a suit challenging the new Florida anti-moderation law
has links to declarations from people involved in moderation at
large online providers including Youtube, Facebook, and Etsy.

https://blog.ericgoldman.org/archives/2021/06/plaintiffs-request-preliminary-injunction-against-floridas-censorship-law-sb-7072-netchoice-v-moody.htm

R's,
John