I'm generally sympathetic to the idea that Section 230 protections should come with some sort of obligation to allow free speech. However, the actual policy proposals for replacing Section 230 are all outright dystopian. Josh Hawley, in particular, is NOT a free speech advocate. His problem with Facebook/Tiwtter is perceived liberal bias, and the alternatives to Section 230 that he suggests are 100% about wrestling e…
I see you chose to attack the person, not the proposal. You are wrong. The bill does not designate a political board, it requires tech companies that have over $30 million U.S. users per month and an annual income of over $1.5 billion, to publish all of their content moderation policies. Users who charge that the companies are not implementing content moderation policies fairly would be able to sue for $5,000 plus at…
Here's Josh's own description of the legislative intent: "Big tech companies would have to prove to the FTC by clear and convincing evidence that their algorithms and content-removal practices are politically neutral. The FTC could not certify big tech companies for immunity except by a supermajority vote"
If the FTC has the authority that Josh wants it to have then it will 100% be politically weaponized by whoever controls the white house at the time of passage (so, Trump, because it'll only pass if R's sweep in 2021). IMO it's quite naive to think otherwise. In general, but also specifically with respect to Trump.
But, assume Trump is this amazingly neutral and high-minded person uninterested in using political power to shape social media narratives. Okay. I have a PhD in machine learning, have tons of experience designing and deploying systems, and I'm pretty up to date on all of the fairness literature. I have No. Fucking. Clue. how I would convince even myself that a content moderation algorithm is "politically neutral".
Even with a clean spec, this seems hard because content moderation algorithms are huge and complex. Wasn't it just a few years ago that a bug in Java's sorting algorithm was found by trying to certify its correctness? Like, bugs live in freaking sorting algorithms of the most popular languages for years and years. Even with a ridiculously clean spec, ...
...and the spec here isn't nearly as clean as "sort the list". The question of what "politically neutral" even means is extraordinarily political. So even if the FTC wasn't explicitly weaponized -- and, dear god, it will be, because the counter-factural here is insane -- the judgements here will still be implicitly political because the spec ("politically neutral") is inherently political.
Proving that a hugely complex ML algorithm is fair simply won't be some sort of apolitical mathematical exercise.
Also, note well: I wasn't even referring to the Ending Support for Internet Censorship Act specifically. But I don't really want to start a debate around this point because that's all beside the point.
I'm curious. What do you think of the proposal made in this thread that people should be able to use their own filters and the big tech cos should be required to implement a clean api for enabling third party filters?
That seems like it solves the "some people don't want to see X" problem in a pretty politically neutral way, but also in a way that acknowledges the difference between the walled-garden network effects web of 2020 and the more decentralized web of 2005.
Seems strictly superior to Josh's proposal of creating a huge incentive to politically weaponize the FTC and giving that almost certainly weaponized body broad authority.
If you think that the "user-chosen filters" solution is not better than Josh's proposal, I'm really interested to hear why.