Twitter announced it would release the source code behind its recommendation algorithm that curates content for users’ “For You” timelines. The code will be open-source, but it will not contain anything related to ad recommendations, nor will it compromise Twitter’s ability to control attempts by bad actors to manipulate its platform. This move aims to increase transparency and protect user privacy and safety.
Twitter published two separate GitHub repositories containing the source code for some of its recommendation algorithms and the machine learning models that underpin them. The tweets that appear on a user’s timeline are picked by a service called Home Mixer, which uses a pipeline to get the best tweets from various recommendation sources, ranks them using a machine learning model, and applies filters to exclude NSFW content, tweets from blocked users, and those already seen.
The ultimate goal of this algorithm is to show users 50% of the most recent and relevant tweets from their followers and the remaining 50% from people outside their network based on what the user finds interesting. This approach ensures that the timeline displays tweets from diverse sources.
Earlier this month, Twitter’s source code and internal tools were publicly leaked on GitHub, prompting the company to take down the leaked code and tools. The company also requested access history information for the leaked code. It used a subpoena to compel GitHub to share the identity of the person who originally published and distributed the leaked Twitter source code.
In conclusion, Twitter’s decision to open-source its recommendation algorithm’s source code is a positive step towards transparency and user safety. By doing so; the company can ensure that bad actors cannot exploit the algorithm to manipulate the platform.