Roboin Blog

X begins testing Collaborative Notes, in which AI writes Community Notes and self-improves through human feedback

Collaborative Note. Al-drafted, updates with community input.
Image:X

As a new initiative for Community Notes, which add background information to posts, X has announced "Collaborative Notes," in which AI and humans work together to create Community Notes.

In Collaborative Notes, in response to users' requests for Community Notes, AI creates a draft, and through human evaluations and suggestions, the AI improves in real time.

Advertisement

When considering updates to Collaborative Notes, the system reviews evaluations and suggestions from Community Notes contributors, ensures the note is as useful as possible, and then determines whether the new version is a meaningful improvement.

If you propose a change to Collaborative Notes and the note is updated, you will receive a notification, and you can check whether your suggestion was adopted and why.

X explains that this will allow the general public and AI to work together to improve the accuracy of information.

At present, Collaborative Notes are generated only in response to requests from Top Writers on English-language posts.

In addition, to propose changes to Collaborative Notes, a Writing Impact of 2 or higher is required, which indicates how often a note has earned a "helpful" status when rated by other users.

Advertisement

As with other parts of Community Notes, X has released Collaborative Notes as open source.

Share this article

Share on XShare on BlueskyShare on MisskeyShare on LINEShare on Threads

Follow us for updates

Adding us to your preferred sources on Google makes it easier to find our articles on Google. Also, be sure to follow us on X and our RSS feed.

Add as a preferred source on GoogleAdd as a preferred source on GoogleX Logo
著者のアイコン画像

I've been using JavaScript more than my native language since birth. I am nowhere and everywhere on the internet.

I build web apps and browser extensions in TypeScript as a web frontend programmer. I released Shadowban Scanner, a tool that detects shadowbans on X, and Restore Link Card, a tool that brings back link cards. Media outlets in Japan and abroad covered both tools. For iGEM 2023, I built the Wiki for Team Japan-United and helped the team win the Grand Prize. On my blog, I cover news about X and social media, test and troubleshoot bugs, and share frontend development insights.