Twitter Releases Code Showing How It Determines What Appears On Your Timeline

Twitter fulfilled its promise to release some of the code that underpins how it filters and presents information to users on its platform. The code, called the “algorithm,” was published on Github and provides insight into the complex process of determining which tweets users see. This algorithm is crucial for de-amplifying content and filtering information based on content quality. It consists of both people users follow and those they do not, and requires the use of machine learning algorithms to narrow down around 500 million daily tweets to a few top posts that are ultimately shown.

The code also deals with “Trust and Safety,” which has been used by Twitter to hide content, such as during Hunter Biden’s laptop scandal before the 2020 election. Engineers revealed that the code only contains logic and machine learning models but not the data that powers it, indicating limited use. Toxic tweets are identified through a “Toxicity Score,” which evaluates keywords dealing with topics like politics, race and insults, without disclosing them. Twitter also tracks user metrics to avoid negatively impacting one group over others. It classifies some topics as blacklisted and sets a “tweet safety level.”

Policy violations are also considered, including ElectionInterference, MisinformationVoting, HackedMaterials, MisinformationCivic, AbusePolicyUkraineCrisisMisinformation, MisinformationMedical and MisinformationGeneral. The code mentions a “misinfo nudge” and incorporates safety labels such as DoNotAmplify, CoordinatedHarmfulActivity, and CivicIntegrityMisinfo, to name a few.

Some conservatives were recently banned for calling attention to a “Trans Day of Vengeance” set to take place on Saturday, although the original far-left accounts were not banned when they promoted the event.