> 4. Far more complicated, change the algorithm to promote positivity and punish negativity.
How do you do that when what constitutes positive or negative is so subjective?
How does twitter make judgemwnts about whether discourse on its site (about a particular subject) is positive ir negative without making value judgements about the subject itself?
Far more complicated is the issue. The problem is even worse than you lay out. How do you deal with backhanded compliments for example. They are inherently positive while really being negative. How about sarcasm which is itself breaking.
I legit don't have an answer though. So I'm going to go with "Tensor flows!"
How do you do that when what constitutes positive or negative is so subjective?
How does twitter make judgemwnts about whether discourse on its site (about a particular subject) is positive ir negative without making value judgements about the subject itself?