the-algorithm
the-algorithm copied to clipboard
twitter
Remove promotion of toxic content?
Am I reading this right?
https://github.com/twitter/the-algorithm/blob/main/home-mixer/server/src/main/scala/com/twitter/home_mixer/functional_component/decorator/HomeTweetTypePredicates.scala#L188
Are you actually pushing tweets to users /because/ they're detected as toxic?!
No, you are not reading this right. Everything in that file is used for tagging tweets with certain features (has the tweet a link, is there anything toxic in it, has it x images etc). That code has nothing to do with actually selecting tweets for the home view.
The code in https://github.com/twitter/the-algorithm-ml/tree/main/projects/home/recap might help you if you are looking for this. However, you won't be able to replicate Twitter's feed 1:1 since some models and information are still missing.
@VIEWVIEWVIEW thanks for clarifying, I still find it a bit concerning that toxicity is being fed into the ML stuff as an input, I suppose you could argue it may be being used to filter OUT overly toxic content depending on what it thinks I want to see, but since we can't really /see/ what the ML does, I don't trust it
I still find it a bit concerning that toxicity is being fed into the ML stuff as an input
how would the model learn about toxic stuff if it isn't fed toxic stuff
how would the model learn about toxic stuff if it isn't fed toxic stuff
As the old saying goes "how would you know that you are dead if the machine already killed you?"
Anyway I think we are digressing a little bit here, we can close this issue as it was fixed already by https://github.com/twitter/the-algorithm/issues/308
