4 ms·

This was a mistake due to this specific blog post sharing some design elements as some spammy blogspot posts and it was fixed. If we really wanted to censor thi

by secalex 11y ago

This was a mistake due to this specific blog post sharing some design elements as some spammy blogspot posts and it was fixed. If we really wanted to censor this story wouldn't we block all criticism of Facebook and not a pretty straightforward and well-reasoned analysis of Safe Harbor? Wouldn't we block the "I declare that I am a wizard and Facebook can't use my content blah blah" memes?

thibaut_barrere 11y ago
Thanks for fixing this. I personally did not assume this was necessarily censorship, but don't be surprised by the reaction of users though, I think you'll see more of these reactions for a while, after the Safe Harbor ruling.
- mkagenius 11y ago
  True. I doubt facebook will do anything that prevents a healthy discussion (unless its unlawful) for its USP is users.
hellofunk 11y ago
Interesting... other blog posts by the same author on the same blog were not blocked, just this one particular post. I find it not a coincidence that the post mentioned Facebook quite a bit, and thus it was blocked. I am doubtful this was due to "design elements" without more details.
rawdisk 11y ago
"... due to sharing some design elements..." But someone downthread said the .co.uk version was not blocked (before you unblocked the .com). Are you implying the .co.uk version did not have the same "design elements"? As to your proposed argument, I think selectively blocking well-reasoned analysis by law professors and letting memes go unblocked makes the most sense for your company.
- amyjess 11y ago
  > Are you implying the .co.uk version did not have the same "design elements"? Classifiers do weird things, and features you don't expect to be significant can suddenly have a much larger than intended effect. The domain name, as a feature, must've thrown it over the edge. It's like that Google+ post where a picture of a couple of black people got tagged as "gorillas" by their auto-tagging ML system. No, Google isn't racist, their classifier just hates their PR department. Ah, the joys of machine learning...
- secalex 11y ago
  Probably because this post was getting lots of links for a domain that is otherwise rarely seen on FB? You can read about the challenges of spam-fighting at scale (and people actually getting paid to write Haskell) here: http://www.wired.com/2015/09/facebooks-new-anti-spam-system-hints-future-coding/ http://www.wired.com/2015/09/facebooks-new-anti-spam-system-...
astrange 11y ago
If people can't tell the difference between your spam filter and a censor, at least you've passed the Turing test.

by secalex 11y ago

thibaut_barrere 11y ago

Thanks for fixing this. I personally did not assume this was necessarily censorship, but don't be surprised by the reaction of users though, I think you'll see more of these reactions for a while, after the Safe Harbor ruling.

mkagenius 11y ago
True. I doubt facebook will do anything that prevents a healthy discussion (unless its unlawful) for its USP is users.

hellofunk 11y ago

Interesting... other blog posts by the same author on the same blog were not blocked, just this one particular post. I find it not a coincidence that the post mentioned Facebook quite a bit, and thus it was blocked. I am doubtful this was due to "design elements" without more details.

rawdisk 11y ago

"... due to sharing some design elements..." But someone downthread said the .co.uk version was not blocked (before you unblocked the .com). Are you implying the .co.uk version did not have the same "design elements"? As to your proposed argument, I think selectively blocking well-reasoned analysis by law professors and letting memes go unblocked makes the most sense for your company.

amyjess 11y ago
> Are you implying the .co.uk version did not have the same "design elements"? Classifiers do weird things, and features you don't expect to be significant can suddenly have a much larger than intended effect. The domain name, as a feature, must've thrown it over the edge. It's like that Google+ post where a picture of a couple of black people got tagged as "gorillas" by their auto-tagging ML system. No, Google isn't racist, their classifier just hates their PR department. Ah, the joys of machine learning...
secalex 11y ago
Probably because this post was getting lots of links for a domain that is otherwise rarely seen on FB? You can read about the challenges of spam-fighting at scale (and people actually getting paid to write Haskell) here: http://www.wired.com/2015/09/facebooks-new-anti-spam-system-hints-future-coding/ http://www.wired.com/2015/09/facebooks-new-anti-spam-system-...

astrange 11y ago

If people can't tell the difference between your spam filter and a censor, at least you've passed the Turing test.