Twitter’s Algorithm Became Open Source
Twitter made their code that selected which tweets appear on users’ timelines available on GitHub. After March 31st, everyone in the world had access to the codes. It also published a blog post clarifying this move. The blog post explained several considerations and steps of the algorithm.
What did they post?
- walk you through their algorithm’s steps
According to The Verge, Twitter’s blog post listed three recommendation stages. First, it gathered Tweets from different sources. Second, it ranked tweets by using a machine learning model. After that, it deleted the parts users had seen, blocked, and considered dangerous for work.
- interesting portion of the code: tracking metrics
Twitter explicitly tracks metrics for users divided into four categories:
"power_user"
"republican"
"democrat"
and
"elon"
Twitter was tracking Elon Musk’s tweets!
Musk responded to it during the Twitter Spaces audio chat after the release of codes. As reported by Mashable, Musk said that he did not hear about such codes before, and "There’s a ton of stupid things’. They would make changes based on suggestions.The readme for the GitHub said, “We invite the community to submit GitHub issues and pull requests for suggestions on improving the recommendation algorithm.”
Did they share everything?
As reported by TechCrunch, Twitter excluded code used to train recommendation algorithms. The release included limited instructions and was aimed primarily at developers. Twitter claimed to have excluded code that could compromise user safety and privacy, but this came after the recent firing of its ethical AI and trust and safety staff.
Overall, the transparency of algorithms’ steps increased users’ trust in Twitter and examined how Twitter's recommendation system works, potentially identifying potential biases and flaws. Although it had not released all algorithms, they helped foster a culture of openness and collaboration in the tech industry, leading to the development of more responsible and ethical algorithms.
Words Count: 306
Comments
Post a Comment