“Negative/hate tweets will be max deboosted and demonetized”

Post by tootie » Fri, Nov 18, 2022, 11:20 PM UTC

phphphphp wrote: post #142

boole1854 wrote: post #105

Many commentators seem to think this is a sudden, recent backpedaling from a previous absolutist view on free speech. However, Musk has been promoting the "freedom of speech" vs "freedom of reach" distinction for a while, at least months before his takeover, including describing how such a policy should be implemented at Twitter. Here's one example quote from June, 2022: “I think there’s this big difference between f…

That analogy doesn’t really fit with the model of Twitter. If we are using the town/time square example, what Elon is describing is the digital version of putting a noise-cancelling box around the people saying “bad” things so that nobody can hear them. A message spreads on Twitter because individuals on Twitter amplify it (with engagement, retweets etc.) which is itself a form of speech: if you say something bad, an…

Post by lliamander » Fri, Nov 18, 2022, 11:24 PM UTC

blitzar wrote: post #165

lliamander wrote: post #157

Earlier quoted context omitted.

I disagree. What the free speech side is arguing is that by trying to restrict "reach" you are trying to prevent interested consumers of that speech from ever hearing it. The prime example is when someone's tweets don't show up in the timelines of subscribers. People who specifically what to hear that speech never get a chance to.

Deplatforming is simply a form of limiting reach, a twitter ban isnt an internet ban

Post by chc » Fri, Nov 18, 2022, 11:25 PM UTC

KerrAvon wrote: post #188

danans wrote: post #170

Earlier quoted context omitted.

Look at the history of banned accounts [1] You don't need polynomial regression to figure out that nobody was being banned for traditional, non-hate and non-violence-inducing conservative or liberal speech. It's mostly harassment, threats, impersonation and disinformation. Nobody has been banned for advocating: lower taxes, higher taxes, cuts to government spending, increased government spending, abortion rights, ant…

There is a joke, I don't know the origin: Conservative: I'm being silenced for my opinions! Me: lowering taxes? C: no, not that opinion M: smaller government? C: no, not that one M: which opinions? C: oh, you know the ones...

Post by nomel » Fri, Nov 18, 2022, 11:27 PM UTC

root_axis wrote: post #220

nomel wrote: post #200

Earlier quoted context omitted.

> but it's also the status quo prior to Elon buying twitter. Wasn't the status quo to ban accounts completely, due to individual tweets, rather than hide individual tweets? This seems much much more permissive.

Twitter has historically utilized a wide range of moderation strategies including deranking, post removals, fact checking blurbs, temporary bans, permanent bans and more.

Post by MichaelCollins » Fri, Nov 18, 2022, 11:29 PM UTC

lrvick wrote: post #30

elgenioso wrote: post #4

Earlier quoted context omitted.

It's impossible to detect negative sentiment purely by text. It's an unsolvable problem.

As someone who specialized in this area for years in social media specifically, I have never seen evidence you can get past 80% accuracy with any known sentiment analysis method. Brute force stats like Naive Bayes is still as good as it gets on platforms like Twitter where speech patterns are too short and irregular for traditional NLP to be very useful.