X said on Thursday it will publish the source code for its “For You” timeline and core ranking engine on GitHub and add a tool showing whether ranking systems affected users’ accounts or posts.
The code will be released under the Apache v2 license and will include model configuration, filters and core ranking system details, expanding X’s earlier open-source effort by roughly 10 to 15 times.
“You’ll get the core ranking code that pulls posts and ranks them for any given user and assembles the feed,” X Vice President of Product Keith Coleman told TechCrunch. He said users will also be able to see systems that filter potentially problematic or rule-violating content, and that some systems, including the ranker and score, can be run outside the company.
“This is this kind of thing that I think people will be fairly shocked that we are releasing,” Coleman said.
X is also rolling out a transparency tool on an “Under the Hood” settings page. Users who have posted 10 or more times in the past month will be able to download aggregate statistics as a JSON file showing whether labels were applied to their account or posts during the past calendar month.
The company said the tool will first be offered to a pilot group of accounts that are at least one year old before a broader rollout.
X said non-technical users could use a large language model to interpret the downloaded data alongside the code published on GitHub.
Before launch, X previewed the codebase to external researchers familiar with recommendation systems, and they were able to run the “score,” the numerical value calculated for each post, outside X, Coleman said.
Developers will be able to submit proposed changes to the repository through pull requests, which X engineers may choose to incorporate into the algorithm.
“That would be amazing to have people submitting code that improves the algorithm. I mean, how cool would it be for the X algorithm to be not just visible to the public, but also, like, by the public?” Coleman said.
X is excluding some systems from the release, including those using Grok to predict whether a post may violate rules. The company said it withheld those tools to prevent bad actors from using them to evade rules and spread spam.
The changes are intended to address concerns about how X’s algorithm affects politics, elections and misinformation. Coleman said X wants the public to be able to assess how posts are distributed, judge whether the system is fair and criticize it if they think it is not.
“That’s the whole goal of this: [the code] can be audited; it can be critiqued. We’re going to listen, and we want to make the system one that people like and trust, and feel is fair,” he said.
Featured image credit
