As a moderation: X will have Community Notes written by KI

Soon there will be “Ai Note Writer”, i.e. Ki written information on articles. Community Notes are used on the platform to correct or classify contributions. They are under a post when the corresponding writers believe that a reference is required. This should bring a kind of self -correction on X and replace moderation. However, the AI ​​notices are not automatically displayed. People operate and control them. A community emergency member is responsible for a AI writer.

Community Notes are a more complex system than it initially sounds. Not everyone can write them, and for the publication, the consent of other authors – they often have to mark the content of the notes as helpful, only then they become visible. It will also be for the AI ​​generated Notes. So it is almost suggestions from a AI to provide contributions with notes. In addition, the AI ​​information should be clearly marked as such.

AI and humans complement each other in the moderation

At the beginning, it says in one X contribution from the Accounts for Community Notesthe AI-Note writer cannot just get started anywhere, but only comment on contributions where these comments are requested. If the first tests are successful, that should change.

X promises faster and more extensive community notes. The system is supposed to improve itself by making the community’s feedback better and better. To confirm this, has X a study available on Arxiv with the MIT and the University of Washington brought out in which scientists have examined to what extent AI can support people in evaluating Community Notes. You call the Reinforcement Learning from Community Feedback (RLCF) – people get feedback from Ki – Vice Versa.

Community Notes based on the model of X has also announced Meta for its platforms. The first field tests are already underway. Mark Zuckerberg hopes that the community’s information would be less biased than fact checkers. He has dissolved contracts with these at least in the USA. In the same train, the moderation rules on Facebook, Instagram and threads were clearly softened. Here, too, the community should be measured itself and less stipulated from above.