X Algorithm Source Code Released for Transparency

Elon Musk’s X releases its algorithm source code, promoting fairness and transparency in post recommendations. Explore how it shapes user engagement.

On X, not all engagement carries the same weight. Elon Musk’s social media firm unveiled its recommendation algorithm source code on Thursday, continuing Musk’s push for transparency on the platform.

The company expressed its objective clearly: “The goal is straightforward—we want people to be able to answer for themselves whether a platform is limiting their reach, whether the system is fair, and why they see particular content,” the company stated in a post on X.

Through this announcement, Musk explained that making X open source was intended to “improve fairness” and gather constructive feedback.

The details embedded in the released code reveal how posts are ranked in a user’s For You timeline. The algorithm, known as “Phoenix,” utilizes a user’s viewing history to predict interactions with posts from non-followed accounts.

By carefully analyzing these predicted interactions, Phoenix assigns weighting scores to posts that dictate their accompanying visibility. For instance, should Phoenix foresee that a user would react negatively to a post, this would result in a significant reduction in the content’s reach. The algorithm categorizes various interactions as follows: a predicted report affects visibility 468 times more negatively than a predicted like, while a “mute” interaction has a negative weight of about 118 times that of a like. Furthermore, a ‘not interested’ action carries a negative impact about 86 times larger than that of a like, whereas a block has a negative weight of around 62 times.

Interestingly, this approach suggests that inflammatory content or “ragebaiting” may not be an effective strategy for growing a following, a notable insight considering X’s turbulent reputation under Musk’s management.

Additionally, the Thursday release included a pilot feature dubbed “Under the Hood,” which provides select users with aggregate information about the labels attached to their posts and accounts, which can influence visibility. However, this feature is currently operational only for a test group of eligible accounts, with plans for broader implementation reliant upon feedback.

This latest revelation is part of Musk’s commitment to increasing the transparency of the X platform. Earlier this year, he described the recommendation algorithm as “dumb” and vowed to enhance the system’s openness. “At least you can see us struggle to make it better in real time and with transparency. No other social media companies do this,” Musk remarked on X.