Data and Society published, in mid November 2018, a 50 page
paper “Content or Context Moderation? Antisanal, Community-Reliant, and
Industrial Approaches”, by Robyn Caplan, at this link (downloadble PDF)
The three basic strategies parse according to the kind of
service. Patreon, Medium, and Vimeo are
said to you antisanal (context-based) approaches; Reddit and Wikipedia use community volunteers;
Facebook and Google (especially YouTube) use industrial approaches with
considerable automation (such as Google’s ContentId).
There are many areas the report doesn’t mention. For one
thing, the upcoming implementation of the European Union’s Copyright Directive
(especially Articles 11 and 13) could increase moderation problems for platforms
even for users in the rest of the western world.
The article discusses Section 230 of the 1996
Telecommunications Act in the US, with a “good Samaritan” policy to allow platforms
to set their own moderation standards. Recently CDA230 has been weakened by FOSTA,
the new ant-trafficking law. Violet Blue’s Engadget article (“Congress JustLegalized Sex Censorship: What to Know”, March 2018) gives a detailed rundown
on the self-protective behavior of many platforms.
Caplan doesn’t mention the parallel DMCA Safe Harbor
for copyright. It does compare how downstream liability works in Germany, where
there is a visitor size threshold and where hate speech is illegal, to the US.
The recent bannings by Patreon (crowdfunding) on a rule
based on “manifest observable behavior” as defined in 2017 by CEO Jack Conte,
seem to be based on an inflexible (rather than context-antisanal) approach to
the use of bad words or slurs. Increasingly platforms are willing to ban for
off-platform behavior (beyond the obvious cases of criminal convictions) and
associations, partly out of fear of the alt-right and of covert and hostile foreign
(especially Russian) influence.
No comments:
Post a Comment