Twitter’s recent release of its algorithm’s underlying code on GitHub, as promised by Elon Musk when he proposed taking over the platform, has not revealed significant insights into how the algorithm actually works, according to experts who have analyzed the code. The code shared by Twitter was a heavily redacted version that omitted important details, including systems related to ads, trust, and safety, as well as the underlying models used to train the algorithm. This has left the most crucial part of the algorithm inscrutable, according to Sol Messing, associate professor at NYU’s Center for Social Media and Politics.
Musk’s original motivation for advocating open sourcing the algorithm was to increase transparency and prevent behind-the-scenes manipulation. However, the code released by Twitter has not shed much light on potential bias or manipulation. It lacks insights into why some tweets may be down-ranked or up-ranked, and it does not provide a comprehensive picture of how different types of engagement are prioritized by the algorithm.
Furthermore, Twitter’s recent changes to its API have made it difficult for researchers to access meaningful amounts of Twitter data, hindering their ability to conduct independent audits of the algorithm. Without proper API access, researchers are unable to thoroughly study the algorithm’s behavior, which limits the effectiveness of open sourcing the code as a means of public oversight.
While there is some new information in the code, such as a “formula” for how different types of engagement are weighted by the algorithm, it is still an incomplete picture without access to actual data. The high cost of obtaining meaningful data for research due to the limitations imposed by Twitter makes it challenging for academics to conduct comprehensive studies on the algorithm’s workings.
In conclusion, Twitter’s release of its algorithm’s code on GitHub has not provided significant insights into how the algorithm actually works, as it was a heavily redacted version that omitted crucial details, and Twitter’s recent API changes have limited researchers’ ability to conduct independent audits. Despite the initial promise of increased transparency, the released code has not fully addressed concerns about bias or manipulation in Twitter’s algorithm, and further efforts are needed to achieve meaningful public oversight.
