{"id":19456,"date":"2022-11-29T00:16:13","date_gmt":"2022-11-29T00:16:13","guid":{"rendered":"https:\/\/mailinvest.blog\/index.php\/2022\/11\/29\/twitter-updates-algorithm-powering-community-notes-the-project-formerly-known-as-birdwatch\/"},"modified":"2022-11-29T00:16:13","modified_gmt":"2022-11-29T00:16:13","slug":"twitter-updates-algorithm-powering-community-notes-the-project-formerly-known-as-birdwatch","status":"publish","type":"post","link":"https:\/\/mailinvest.blog\/index.php\/2022\/11\/29\/twitter-updates-algorithm-powering-community-notes-the-project-formerly-known-as-birdwatch\/","title":{"rendered":"Twitter Updates Algorithm Powering Community Notes, the Project Formerly Known as Birdwatch"},"content":{"rendered":"

<\/a>\n
<\/p>\n

\n

Twitter detailed updates to the algorithm powering crowdsourced fact-checking initiative Community Notes, which was known as Birdwatch<\/a> in the pre-Elon Musk<\/a> days.<\/p>\n

Community Notes may take on more importance at the platform, as Twitter\u2019s content moderation teams were gutted as part of mass layoffs<\/a> early this month, followed by a purging of several contractor positions in the middle of the month.<\/p>\n

Twitter said in a series of tweets from the Community Notes account<\/a>, \u201cOne of contributors\u2019 top frustrations is seeing and rating low-quality notes. We just rolled out an algorithm update that identifies more low-quality notes, saving contributors time and improving the average quality of notes (and note writers).\u201d<\/p>\n

\n
\n
\n

One of contributors\u2019 top frustrations is seeing and rating low quality notes. We just rolled out an algorithm update that identifies more low quality notes \u2014 saving contributors time, and improving the average quality of notes (and note writers). How it works\u2026<\/p>\n

\u2014 Community Notes (@CommunityNotes) November 26, 2022<\/a><\/p>\n<\/blockquote><\/div>\n<\/figure>\n

The company explained that while contributors were able to write notes proposing that context be added to a tweet or explaining why a tweet wasn\u2019t misleading and didn\u2019t need the addition of context, the latter type of note generated \u201cnoisy\u201d rating data, so Twitter paused scoring those notes.<\/p>\n

Its new algorithm update resumed scoring those notes with the aim of identifying more low-quality notes, taking away the ability to write notes from people consistently contributing low-quality notes and improving contributor helpfulness scores with more data.<\/p>\n

In response to a question about how the algorithm determined low-quality notes that brought up the potential for censorship, Twitter vice president of product Keith Coleman<\/a> replied in a tweet<\/a>, \u201cThat is an important question. In the case of Community Notes, \u2018low-quality\u2019 equals notes that are found not helpful by a wide range of people, including people who typically disagree with each other. This prevents one-sided outcomes.\u201d<\/p>\n

Twitter introduced the Birdwatch program<\/a> in January 2021, inviting contributors to flag potentially misleading information in tweets and write notes providing context.<\/p>\n

In September, the social network detailed a new onboarding process<\/a> that assigned new contributors who meet its eligibility criteria\u2014a verified phone number from a trusted carrier based in the U.S., as well as a minimum of six months on Twitter, with no recent violations\u2014with an initial Rating Impact score of zero, which they could increase by consistently rating other contributors\u2019 notes and reliably identifying those that are helpful and not helpful.<\/p>\n

They could begin writing notes once their Rating Impact score reached five, and Twitter explained that they could continue to boost that score by writing and rating notes, but consistently writing notes that other contributors rated as not helpful would lower their score and potentially cause their ability to write notes to be temporarily locked.<\/p>\n

Twitter had roughly 15,000 Birdwatch contributors at the time, and it said it would begin adding about another 1,000 per week<\/p>\n

In October, shortly before Musk\u2019s takeover of the company, Twitter said users in the U.S. would begin seeing some tweets accompanied by notes<\/a> containing relevant information about that tweet that was rated as \u201cHelpful\u201d by Birdwatch contributors, adding that most of those notes would provide additional sources for people who want to dive deeper into the subject matter.<\/p>\n

The company said at the time that users would be able to rate the notes they see in order to help Twitter understand whether or not they are helpful.<\/p>\n<\/p><\/div>\n