elgenioso

Even non-traditional NLP won't work. There is no way to algorithmically enforce "good" speech. I agree that the character limitations don't help and the 280 character limitation is the main reason Twitter is such a mess.

There are no coherent long form texts generated by AI models but 280 characters is very easy because people will give the author the benefit of the doubt and interpolate the missing context.

This is not a solvable problem. There is no algorithmic way to detect speech that needs to be boosted and deboosted on social media because there is no way to actually algorithmically detect what is truthful and what is just manipulative propaganda and marketing.

I don't think that's true. The only way to limit speech on an algorithmicially mediated platform is via algorithms and there is no algorithm that can encode non-trivial properties of language regardless of what the AI maximalists might say to the contrary.

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